FR3043233A1 - METHOD OF AUTOMATICALLY SELECTING IMAGES FROM A MOBILE DEVICE - Google Patents

Abstract

The present invention relates to a method for automatic selection of at least one original digital image (10) among at least a plurality of images, said at least one original image (10) being defined by a matrix of pixels, said method being configured to be executed by at least one processor included in at least one mobile device, and said method being characterized in that it comprises at least: a plurality of steps for analyzing the at least one original image; A step of generating at least one miniature image (11) being formed by a miniature pixel matrix constituted by a portion of said pixel matrix of the original image (10); At least one of said plurality of analysis steps configured to use said thumbnail image (11); A selection step based on said overall score of the original image (10);

Description

TECHNICAL FIELD OF THE INVENTION The invention relates to the field of image processing, and more particularly to the processing of the image on a mobile device. It finds for example particularly advantageous application of the field of smart phones, usually referred to by their English word "smartphones", and which are equipped with an optics for taking pictures.

STATE OF THE ART

Photography has never known so many practitioners in all of its history. Today more than ever anyone can capture his environment at will and make thousands of photos.

Thanks to smartphones, so many photos have been taken in the last three years as in the history of photography. But this massive shooting of photos around the world poses multiple problems. The modern amateur photographer is confronted with new problems that are the storage and selection of the best shots among a multitude.

Currently many solutions exist that allow to process according to specific criteria dozens of photos so as to isolate the most relevant in view of these criteria. These tools, first designed for professional photographers who can make hundreds of shots at events, are now within everyone's reach. However, these tools of great power require just a lot of computing power. Thus it is difficult, if not impossible, to operate these tools on mobile devices like smartphones in a reasonable time. The known solutions proposed are therefore the transfer of photos to servers or processing units relocated. This involves having a network connection and also the transfer of private photo for example. To these drawbacks, new and recent solutions are known that make it possible to process images from the smartphone itself locally. However, the processing time is substantial, system resources and energy autonomy are put to the test.

In view of these drawbacks, the present invention proposes a solution for partially solving at least the disadvantages previously presented.

SUMMARY OF THE INVENTION

The present invention relates to a method of automatically selecting at least one digital original image from at least a plurality of digital original images, said at least one original image being defined by a matrix of pixels, said matrix being defined by at least one a central pixel area and at least one outer pixel area surrounding said central pixel area, each pixel being defined by at least one digital brightness and saturation value, said method being configured to be executed by at least one processor included in at least one mobile device comprising at least one non-transient memory on which the plurality of original images is recorded, at least one power source, at least one digital image pickup sensor and at least one display screen , and said method being characterized in that it comprises at least: - A plurality of steps of analyzing the at least one original image, said method being configured to assign a global evaluation score to the original image, the overall evaluation score being a numerical value configured to be incremented by at least one score partial evaluation assigned to the original image during at least one of said plurality of analysis steps; A step of generating at least one miniature image being formed by a miniature pixel matrix constituted by only a portion of said pixel matrix of the original image; At least one of said plurality of analysis steps configured to use said thumbnail image so as to attribute to said original image the at least one partial evaluation score; A selection step based on said overall score of the original image;

The present invention allows automatic selection of the best images from an image library. This selection is based on relevant robust analysis criteria. In addition this selection is made on a mobile device type smartphone for example.

The present invention is particularly optimized to operate on a mobile device so as to save the system resources of the mobile device but also its energy source. Indeed, the present invention is designed so that on the one hand the image analysis steps are low in computing power and therefore low in energy consumption, and on the other hand the available options allow a considerable gain of energy. memory space by removing images of poor quality or being taken several times in a row.

The present invention does not require any connection to a communication network, which therefore allows energy saving and data flow. The use of miniature in the present invention ensures both an increased speed of the automatic selection method but also a saving of the battery of the mobile device since the CPU cycles are reduced because the analysis algorithms only work on a part of the image and not on all the pixels that it understands. This trick thus saves the system and energy resources of the mobile device. This trick surprisingly presents a very good quality of relevance in the selection of better quality images.

Finally, the present invention adapts the images to the automatic selection method designed to consume few system resources instead of adapting system resources to the images.

The present invention also relates to a computer program product stored in a non-transitory memory and having instructions executable by at least one processor, configured to execute the method according to the present invention.

The present invention also relates to a mobile device comprising at least one optical digital imaging sensor, typically a digital camera, at least one processor, at least one power supply source and at least one non-transitory memory configured to back up. the images taken by the sensor and at least one memory storing a computer program product comprising instructions executable by said at least one processor, said instructions being configured to execute the method according to the present invention.

Other objects, features and advantages of the present invention will become apparent from the following description and accompanying drawings. It is understood that other benefits may be incorporated.

BRIEF DESCRIPTION OF THE FIGURES

The objects, objects, as well as the features and advantages of the invention will become more apparent from the detailed description of an embodiment of the latter which is illustrated by the following accompanying drawings in which: FIG. 1 represents a diagram of the sequence of the different steps of the present invention according to one embodiment. FIGURE 2 is a schematic diagram of the step of generating a thumbnail image from an original image according to an embodiment of the present invention.

The drawings are given by way of examples and are not limiting of the invention. They constitute schematic representations of principle intended to facilitate the understanding of the invention.

DETAILED DESCRIPTION OF THE INVENTION

Before starting a detailed review of embodiments of the invention, are set forth below optional features that may optionally be used in combination or alternatively:

Advantageously, the step of generating the at least one miniature image of the at least one original image comprises a step of duplicating the at least one original image so as to create at least one duplicated image, said at least one duplicated image comprising a plurality of pixels forming a pixel array and a step of removing at least a plurality of pixels from the pixel array of the duplicated image to form the at least one thumbnail image.

This makes it possible to simply and efficiently make a thumbnail image of the original image by removing the pixels that are not relevant for the analysis of the image.

Advantageously, the pixel array of the original image is defined by at least one central pixel area and at least one outer pixel area surrounding said central pixel area and wherein the step of removing said at least one plurality of pixels pixels is configured so that the thumbnail image is formed by a square miniature pixel array.

The square shape allows optimization of the algorithms by working directly with a matrix of square pixels.

Advantageously, the size of the miniature pixel matrix is between 75 pixels by 75 pixels and 1200 pixels by 1200 pixels, preferably between 150 pixels by 150 pixels and 600 pixels by 600 pixels, and preferably equal to 300 pixels by 300 pixels. pixels.

These dimensions make it possible, surprisingly, to select the best photos with great pertinence.

Advantageously, the pixel array of the original image is defined by at least one central pixel area and at least one outer pixel area surrounding said central pixel area and in which the thumbnail generation step is configured. so that at least a portion of the central pixel area of the original image is retained in the miniature pixel array of the thumbnail image.

This makes it possible to simply and efficiently make a thumbnail image of the original image by removing the pixels belonging to the edge of the image. Indeed the edges of the image do not generally include important information, it is then more effective to work only on the center of the image.

Advantageously, the pixel array of the original image is defined by at least one central pixel area and at least one outer pixel area surrounding said central pixel area.

Advantageously, the step of removing at least a plurality of pixels from the pixel matrix of the duplicated image is configured to symmetrically remove the at least one outer pixel area from the pixel area. Central.

Advantageously, the step of deleting at least a plurality of pixels from the pixel array of the duplicated image is configured to symmetrically suppress bands of vertical and / or horizontal pixels on either side of the image. the central pixel area.

Advantageously, the step of removing at least a plurality of pixels from the pixel array of the duplicate image is configured to remove a periphery of pixels surrounding the central pixel area.

This makes it possible to simply and efficiently make a thumbnail image of the original image by removing the pixels belonging to the edge of the image. Indeed the edges of the image do not generally include important information, it is then more effective to work only on the center of the image.

Advantageously, the step of generating the miniature image is configured so that at least part of the at least one outer pixel area of the original image is stored in the miniature pixel array of the image. miniature image.

This makes it simple and efficient to create a miniature image of the original image. Taking into account the edges of the image can sometimes be useful depending on the type of photo taken.

Advantageously, the step of generating the miniature image is configured so that at least a portion of the central pixel area and the at least one outer pixel area of the original image are retained in the image. miniature pixel matrix of the thumbnail image.

This makes it possible to simply and efficiently perform a thumbnail image of the original image while taking into account a specific area of the image according to its content and not only by limiting itself to the center of it.

Advantageously, at least one partial evaluation score is assigned to the at least one original image during at least one of the plurality of analysis steps.

This makes it possible to select the photos according to their overall score which is then calculated according to a plurality of partial scores allocated during the various stages of the automatic selection.

Advantageously, the at least one partial evaluation note is a numerical value.

This makes it possible to select the photos based on their overall rating score which is then calculated based on a plurality of partial scores assigned during the various stages of the automatic selection.

Advantageously, the at least one overall evaluation score is calculated from the at least one partial evaluation score.

This makes it possible to select the photos based on their overall rating score which is then calculated based on a plurality of partial scores assigned during the various stages of the automatic selection.

Advantageously, the at least one plurality of steps for analyzing the at least one original image comprises at least one step of analyzing the resolution of the original image comprising an evaluation of the number of pixels of the at least one an original image, and in which the overall score of the original image is set to zero when the number of pixels of the at least one original image is less than a resolution threshold, said resolution threshold being comprised between 250 000 and 4 000 000 pixels, preferably between 500 000 and 2 000 000 pixels, and preferably equal to 1 000 000 pixels.

This eliminates photos whose resolution is too low.

Advantageously, the at least one plurality of analysis steps of the at least one original image comprises at least one step of evaluating the number of pixels of the at least one original image.

This eliminates photos whose resolution is too low.

Advantageously, the overall evaluation score of the original image is set to zero when the number of pixels of the at least one original image is less than a resolution threshold, said resolution threshold being between 250,000 and 4. 000 000, preferably between 500 000 and 2 000 000, and preferably equal to 1 000 000.

This eliminates photos with too low resolution

Advantageously, the overall evaluation score of the original image is set to zero when the number of pixels of the at least one original image is less than a resolution threshold.

This eliminates photos whose resolution is too low.

Advantageously, the resolution threshold is between 250,000 and 4,000,000, preferably between 500,000 and 2,000,000, and preferably equal to 1,000,000.

This eliminates photos whose resolution is too low.

Advantageously, the resolution threshold is a percentage of the number of pixels of the at least one original image, said percentage being between 10% and 90%, preferably between 15% and 50% and preferably equal to 25%.

This eliminates photos whose resolution is too low while adapting to the type of image considered.

Advantageously, the at least one plurality of steps for analyzing the at least one original image comprises at least one step of evaluating an average brightness of the miniature image, said average brightness of the miniature image being calculated by averaging the luminosities of each pixel of the miniature pixel array, and wherein the overall evaluation score of the original image is set to zero when the average brightness of the thumbnail is greater than a threshold value of maximum brightness, said threshold value of maximum brightness being between 25% and 95%, preferably between 45% and 85%, and preferably equal to 70%, and / or when the average brightness of the miniature image is less than a threshold value of minimum brightness, said threshold value of minimum brightness being between 5% and 55%, advantageously between 15% and 35%, and preferably equal to 20%.

This allows you to select images whose brightness is in a suitable range.

This eliminates images on exposed and / or underexposed. The use of thumbnail greatly reduces the number of pixels to be considered in this evaluation.

Advantageously, the at least one plurality of steps of the at least one original image comprises at least one step of evaluating an average brightness of the miniature image, said average brightness of the miniature image being calculated by realizing the average of the luminosities of each pixel of the miniature pixel matrix of the miniature image. This allows you to select images whose brightness is in a suitable range.

This eliminates images on exposed and / or underexposed. The use of thumbnail greatly reduces the number of pixels to be considered in this evaluation.

Advantageously, the at least one plurality of analysis steps of the at least one original image comprises at least one step of evaluating an average brightness of the miniature image.

This allows you to select images whose brightness is in a suitable range.

This eliminates images on exposed and / or underexposed. The use of thumbnail greatly reduces the number of pixels to be considered in this evaluation.

According to one embodiment, the memory of the mobile device contains said at least one original image.

Advantageously, the average brightness of the miniature image is calculated by averaging the luminosities of each pixel of the miniature pixel array of the thumbnail image.

This allows you to select images whose brightness is in a suitable range.

This eliminates images on exposed and / or underexposed. The use of thumbnail greatly reduces the number of pixels to be considered in this evaluation.

Advantageously, the at least one plurality of analysis steps of the at least one original image comprises at least one step of evaluating an average brightness of the original image.

This allows you to select images whose brightness is in a suitable range.

This eliminates images on exposed and / or underexposed.

Advantageously, the overall evaluation score of the original image is set to zero when the average brightness of the miniature image is greater than a maximum brightness threshold value, said maximum brightness threshold value being between 25% and 95%. , preferably between 45% and 85%, and preferably equal to 70% and when the average brightness of the miniature image is less than a minimum brightness threshold value, said threshold value of minimum brightness being between 5% and 55% , advantageously between 15% and 35%, and preferably equal to 20%.

This allows you to select images whose brightness is in a suitable range.

This eliminates images on exposed and / or underexposed.

Advantageously, the overall evaluation score of the original image is set to zero when the average brightness of the thumbnail image is greater than a maximum brightness threshold value, said maximum brightness threshold value is between 25% and 95% preferably between 45% and 85%, and preferably equal to 70%.

This allows you to select images whose brightness is in a suitable range.

This eliminates images on exposed and / or underexposed.

Advantageously, the overall evaluation score of the original image is set to zero when the average brightness of the thumbnail image is greater than a threshold value of maximum brightness.

This eliminates images on exposed and / or underexposed. The use of thumbnail greatly reduces the number of pixels to be considered in this evaluation.

Advantageously, the overall evaluation score of the original image is set to zero when the average brightness of the original image is greater than a threshold value of maximum brightness.

This eliminates images on exposed and / or underexposed.

Advantageously, the threshold value of maximum brightness is between 25% and 95%, preferably between 45% and 85%, and preferably equal to 70%.

This allows you to select images whose brightness is in a suitable range.

This eliminates images on exposed and / or underexposed.

Advantageously, the overall evaluation score of the original image is set to zero when the average brightness of the miniature image is less than a minimum brightness threshold value, said minimum brightness threshold value is between 5% and 55%. , advantageously between 15% and 35%, and preferably equal to 20%.

This allows you to select images whose brightness is in a suitable range.

This eliminates images on exposed and / or underexposed.

Advantageously, the overall evaluation score of the original image is set to zero when the average brightness of the thumbnail image is less than a threshold value of minimum brightness.

This eliminates images on exposed and / or underexposed. The use of thumbnail greatly reduces the number of pixels to be considered in this evaluation.

Advantageously, the overall evaluation score of the original image is set to zero when the average brightness of the original image is less than a threshold value of minimum brightness.

This eliminates images on exposed and / or underexposed.

Advantageously, the threshold value of minimum brightness is between 5% and 55%, advantageously between 15% and 35%, and preferably equal to 20%.

This allows you to select images whose brightness is in a suitable range.

This eliminates images on exposed and / or underexposed.

Advantageously, the at least one plurality of analysis steps of the at least one original image comprises at least one step of evaluating the sharpness of the miniature image.

This eliminates images that are fuzzy or have insufficient sharpness. The use of thumbnail greatly reduces the number of pixels to be considered in this evaluation.

Advantageously, the at least one plurality of analysis steps of the at least one original image comprises at least one step of evaluating the sharpness of the original image.

This eliminates images that are fuzzy or have insufficient sharpness.

Advantageously, the at least one sharpness evaluation step is based on an edge detection technique, a contour being defined as a set of pixels delimiting two continuous zones of pixels, and in which the percentage of pixels, relative to to the miniature pixel matrix, belonging to a contour is evaluated, and in which the overall evaluation score of the original image is set to zero when said percentage of pixels is less than a minimum contour threshold value, said threshold value of minimum contour being between 0.5% and 4.5%, preferably between 1.5% and 3.5%, and preferably equal to 2.5% and in which the overall evaluation score is incremented by a partial evaluation score of the step for evaluating the sharpness attributed to the original image, said partial evaluation score of the sharpness evaluation step being a function of said pixel percentage s and preferably being between 0 and 1, and preferably between 0.1 and 0.5.

This eliminates images that are fuzzy or have insufficient sharpness.

The edge detection technique is both very fast and very reliable and takes advantage here of the reduced number of pixels to consider.

Advantageously, the at least one sharpness evaluation step uses an edge detection technique, a contour being defined as a set of pixels limiting two groups of pixels, and in which the percentage of pixels, relative to the matrix. of miniature pixels, belonging to an outline is evaluated.

This eliminates images that are fuzzy or have insufficient sharpness.

The edge detection technique is both very fast and very reliable and takes advantage here of the reduced number of pixels to consider.

Advantageously, the overall evaluation score of the original image is set to zero when the percentage of pixels, relative to the miniature pixel matrix, belonging to a contour is less than a minimum contour threshold value, said threshold value of minimum contour is between 0.5% and 4.5%, preferably between 1.5% and 3.5%, and preferably equal to 2.5%.

This eliminates images that are fuzzy or have insufficient sharpness.

Advantageously, the overall evaluation score of the original image (10) is set to zero when the percentage of pixels, relative to the miniature pixel matrix, belonging to a contour is less than a minimum contour threshold value.

This eliminates images that are fuzzy or have insufficient sharpness.

Advantageously, the minimum contour threshold value is between 0.5% and 4.5%, preferably between 1.5% and 3.5%, and preferably equal to 2.5%.

This eliminates images that are fuzzy or have insufficient sharpness.

Advantageously, the overall evaluation score is incremented by a partial evaluation score of the sharpness evaluation step attributed to the original image, said partial evaluation score of the evaluation step of the evaluation. sharpness having no limit to its numerical value.

Advantageously, the overall evaluation score is incremented by a partial evaluation score of the sharpness evaluation step attributed to the original image, said partial evaluation score of the evaluation step of the evaluation. sharpness being between 0 and 1, and preferably between 0.1 and 0.5.

This makes it possible to value the images according to their richness and their sharpness. Indeed, the more contour and richer the image is in terms of details.

Advantageously, the overall evaluation score is incremented by a partial evaluation score of the sharpness evaluation step is attributed to the original image.

This makes it possible to value the images according to their richness and their sharpness. Indeed, the more contour and richer the image is in terms of details.

Advantageously, the partial evaluation score of the sharpness evaluation step is between 0 and 1, and preferably between 0.1 and 0.5.

Advantageously, the partial evaluation score of the sharpness evaluation step is between 0 and 1, and preferably between 0.1 and 0.5.

This makes it possible to value the images according to their richness and their sharpness. Indeed, the more contours and the richer the image is in terms of details. Setting a limit on this note makes it possible to standardize the analysis steps between them in the selection process.

Advantageously, the partial evaluation score of the sharpness evaluation step is doubled so that the partial evaluation score is less than or equal to a threshold value of the partial evaluation score of the step evaluation of the sharpness.

This makes it possible to value the images according to their richness and their sharpness. Indeed, the more contour and richer the image is in terms of details. Setting a limit on this note makes it possible to standardize the analysis steps between them in the selection process.

Advantageously, the at least one plurality of analysis steps of the at least one original image comprises at least one additional analysis step of at least a plurality of pixels composing the miniature pixel matrix of the miniature image. , said additional analysis step being configured to evaluate the brightness and / or saturation of each pixel composing the miniature pixel array, and wherein the percentage of pixels of the miniature pixel array, whose brightness is greater than a value saturated brightness threshold and / or whose saturation is greater than a saturation threshold value, is evaluated so as to increment the overall evaluation score of the original image by a partial evaluation score corresponding to said percentage of pixels, said threshold value of saturated brightness being between 25% and 75%, preferably between 40% and 60% and preferably equal to 50%, and said val saturation threshold being between 25% and 75%, preferably between 40% and 60% and preferably equal to 50%, and said evaluation partial score being between 0 and 1 and preferably between 0.45 and 0.65.

This reinforces the automatic selection process by highlighting the properties of each pixel composing the thumbnail image. This step again benefits from the reduced number of pixels to consider.

Advantageously, the at least one plurality of analysis steps of the at least one original image comprises at least one additional analysis step of at least a plurality of pixels composing the miniature pixel matrix of the miniature image. .

This reinforces the automatic selection process by highlighting the properties of each pixel composing the thumbnail image. This step again benefits from the reduced number of pixels to consider.

Advantageously, the at least one additional analysis step is configured to evaluate the brightness of each pixel composing said at least one plurality of pixels composing the miniature pixel matrix of the miniature image, and to evaluate the saturation of each component pixel. said at least one plurality of pixels composing the miniature pixel array of the thumbnail image.

This reinforces the automatic selection process by highlighting the properties of each pixel composing the thumbnail image. This step again benefits from the reduced number of pixels to consider.

Advantageously, the at least one additional analysis step is configured to evaluate the brightness of each pixel composing said at least one plurality of pixels composing the miniature pixel matrix of the thumbnail image.

This reinforces the automatic selection process by highlighting the properties of each pixel composing the thumbnail image. This step again benefits from the reduced number of pixels to consider.

Advantageously, the percentage of pixels whose brightness is greater than a saturated brightness threshold value and whose saturation is greater than a saturation threshold value, is evaluated so as to increment the overall evaluation score by a partial score of evaluation to the original image corresponding to said percentage of pixels, said threshold value of saturated brightness being between 25% and 75%, preferably between 40% and 60% and preferably equal to 50%, and said threshold value of saturation being between between 25% and 75%, preferably between 40% and 60% and preferably equal to 50%.

This reinforces the automatic selection process by highlighting the properties of each pixel composing the thumbnail image. This step again benefits from the reduced number of pixels to consider.

Advantageously, the number of pixels constituting said at least one plurality of pixels composing the miniature pixel matrix of the miniature image and whose brightness is greater than a saturated brightness threshold value, is evaluated, said saturated brightness threshold value. is between 25% and 75%, preferably between 40% and 60% and preferably equal to 50%.

This reinforces the automatic selection process by highlighting the properties of each pixel composing the thumbnail image. This step again benefits from the reduced number of pixels to consider.

Advantageously, the number of pixels constituting said at least one plurality of pixels constituting the miniature pixel matrix of the miniature image and whose saturation is greater than a saturation threshold value, is evaluated, said saturated brightness threshold value is between 25% and 75%, preferably between 40% and 60% and preferably equal to 50%.

This reinforces the automatic selection process by highlighting the properties of each pixel composing the thumbnail image. This step again benefits from the reduced number of pixels to consider.

Advantageously, the number of pixels constituting said at least one plurality of pixels composing the miniature pixel matrix of the miniature image and whose brightness is greater than a saturated brightness threshold value, is evaluated.

This reinforces the automatic selection process by highlighting the properties of each pixel composing the thumbnail image. This step again benefits from the reduced number of pixels to consider.

Advantageously, the threshold value of saturated brightness is between 25% and 75%, preferably between 40% and 60% and preferably equal to 50%.

This allows you to select photos based on their visual dynamics.

Advantageously, the at least one analysis step is configured to evaluate the saturation of each pixel composing said at least one plurality of pixels composing the miniature pixel matrix of the miniature image.

This allows you to select photos based on their visual dynamics.

Advantageously, the at least one additional analysis step is configured to evaluate the saturation of each pixel composing said at least one plurality of pixels composing the matrix of pixels of the original image.

This allows you to select photos based on their visual dynamics.

Advantageously, the number of pixels constituting said at least one plurality of pixels composing the miniature pixel matrix of the miniature image and whose saturation is greater than a saturation threshold value, is evaluated.

This allows you to select photos based on their visual dynamics.

Advantageously, the saturation threshold value is between 25% and 75%, preferably between 40% and 60% and preferably equal to 50%.

This allows you to select photos based on their visual dynamics.

Advantageously, the percentage of pixels, with respect to the miniature pixel matrix, whose brightness is greater than a saturated brightness threshold value and whose saturation is greater than a saturation threshold value, is evaluated so as to increment the overall score of evaluation by a partial evaluation note to the original image corresponding to said percentage of pixels.

This allows you to select photos based on their visual dynamics.

Advantageously, the partial evaluation score of the at least one analysis step is between 0 and 1 and preferably between 0.45 and 0.65. This allows you to select photos based on their visual dynamics.

Advantageously, the at least one plurality of analysis steps of the at least one original image comprises at least one step of detecting faces present in the original image, and in which for each detected face, the overall score of evaluation is incremented by a partial evaluation score attributed to the original image, said partial evaluation score being a function of the number of faces detected and being 0.05 and 0.5, advantageously between 0.1 and 0.25, and preferably equal to 0.15.

This makes it possible to value the photos according to their content such as the number of people.

Advantageously, the at least one plurality of analysis steps of the at least one original image comprises at least one step of detecting faces present in the original image.

This makes it possible to value the photos according to their content such as the number of people.

Advantageously, for each detected face, the overall evaluation score is incremented by a partial evaluation score attributed to the original image, said partial evaluation score being between 0.05 and 0.5, advantageously between 0.1 and 0.25, and preferably equal to 0.15.

This makes it possible to value the photos according to their content such as the number of people.

Advantageously, for each detected face, the overall evaluation score is incremented by a partial evaluation score attributed to the original image.

This makes it possible to value the photos according to their content such as the number of people.

Advantageously, the partial evaluation score assigned to the original image for each detected face is between 0.05 and 0.5, advantageously between 0.1 and 0.25, and preferably equal to 0.15.

This makes it possible to value the photos according to their content such as the number of people.

Advantageously, the at least one face detection step comprises a smile detection step.

This allows you to value photos based on their content, such as the number of people with a smile.

Advantageously, for each smile detected, the overall evaluation score is incremented by a partial evaluation score attributed to the original image.

This allows you to value photos based on their content, such as the number of people with a smile.

Advantageously, the partial evaluation score attributed to the original image for each detected smile is between 0.1 and 0.5, advantageously between 0.15 and 0.25, and preferably equal to 0.2.

This allows you to value photos based on their content, such as the number of people with a smile.

Advantageously, the at least one plurality of analysis steps of the at least one original image comprises at least one duplicate detection step among the at least one plurality of original images.

This makes it possible to value the photos taken several times indicating the importance of the photo, while selecting the best image.

Advantageously, the duplicate detection step is configured to detect a duplicate when the average luminosities of two miniature images have a difference between them less than a doublon threshold value, said duplicate threshold value being between 0 and 0.4, preferably between 0.05 and 0.2, and preferably equal to 0.1, and when the instants of the recording on the non-transitory memory of said two original images have a temporal difference between them less than a threshold value of time, said threshold value of time being between 0 seconds and 30 seconds, preferably between 5 seconds and 20 seconds, and preferably equal to 10 seconds, and when the percentages of pixels belonging to one contour of said two miniature images have a difference between them less than a threshold value of details, said threshold value of details is between 0 and 0.12, preferably between 0.015 and 0.06, and preferably equal to 0.03, and wherein the step of detecting duplicates comprises a step of selecting an original image whose overall evaluation score is the highest among the detected duplicates for this original image at the time of the duplicate detection step, the overall evaluation score of the selected original image is incremented by a partial evaluation score, said partial evaluation score being between 0 and 0.1 and being multiplied by the number of duplicates detected.

This makes it possible to value the photos taken several times indicating the importance of the photo over a precise time range in order to consider only the images resulting from a repeated take of the same content.

Advantageously, a duplicate is detected when the average luminosities of two miniature images have a difference between them less than a doublon threshold value, said doublet threshold value being between 0 and 0.4, preferably between 0.05 and 0.2, and preferably equal to 0.1, and when the instants of the recording on the non-transitory memory of said two original images (10) have a temporal difference between them less than a threshold value of time, said threshold value of time being between 0 seconds and 30 seconds, of preferably between 5 seconds and 20 seconds, and preferably equal to 10 seconds, and when the percentages of pixels belonging to a contour of said two miniature images have a difference between them less than a threshold value of details, said threshold value of details is between 0 and 0.12, preferably between 0.015 and 0.06, and preferably Equally equal to 0.03 This makes it possible to value the photos taken several times indicating the importance of the photo over a precise time range in order to consider only the images resulting from a repeated take of the same content.

Advantageously, a duplicate is detected when the average luminosities of two miniature images have a difference between them less than a doublon threshold value, said doublet threshold value being between 0 and 0.4, preferably between 0.05 and 0.2, and preferably equal to 0.1, and when the instants of the recording on the non-transitory memory of said two original images (10) have a temporal difference between them less than a threshold value of time, said threshold value of time being between 0 seconds and 30 seconds, of preferably between 5 seconds and 20 seconds, and preferably equal to 10 seconds.

This makes it possible to value the photos taken several times indicating the importance of the photo over a precise time range in order to consider only the images resulting from a repeated take of the same content.

Advantageously, a duplicate is detected when the average luminosities of two miniature images have a difference between them less than a doublon threshold value, said doublet threshold value being between 0 and 0.4, preferably between 0.05 and 0.2, and preferably equal to 0.1, and when the percentages of pixels belonging to a contour of said two miniature images have a difference between them less than a threshold value of details, said threshold value of details is between 0 and 0.12, preferably between 0.015 and 0.06, and preferentially equal to 0.03 This makes it possible to value the photos taken several times indicating the importance of the photo over a precise time range in order to consider only the images resulting from a repeated take of the same content.

Advantageously, a duplicate is detected when the instants of the recording on the non-transitory memory of said two original images have a time difference between them less than a threshold value of time, said threshold value of time being between 0 seconds and 30 seconds, of preferably between 5 seconds and 20 seconds, and preferably equal to 10 seconds, and when the percentages of pixels belonging to a contour of said two miniature images have a difference between them less than a threshold value of details, said threshold value of details is between 0 and 0.12, preferably between 0.015 and 0.06, and preferably equal to 0.03. This makes it possible to value the photos taken several times indicating the importance of the photo over a precise time range in order to consider only the images resulting from a repeated take of the same content.

Advantageously, the duplicate detection step is performed from the average brightness of each miniature image of the at least one plurality of original images and whose instants of the recording on the non-transitory memory present a time difference between them less than a threshold value of time.

This makes it possible to value the photos taken several times indicating the importance of the photo over a precise time range in order to consider only the images resulting from a repeated take of the same content.

Advantageously, the duplicate detection step is performed from the average brightness of each miniature image of the at least one plurality of original images and whose percentages of pixels belonging to a contour of said two miniature images show a difference between them less than a threshold value of details.

This makes it possible to value the photos taken several times indicating the importance of the photo over a precise time range in order to consider only the images resulting from a repeated take of the same content.

Advantageously, the threshold value of time is between 0 seconds and 30 seconds, preferably between 5 seconds and 20 seconds, and preferably equal to 10 seconds.

This makes it possible to value the photos taken several times indicating the importance of the photo over a precise time range in order to consider only the images resulting from a repeated take of the same content.

Advantageously, at least one duplicate is detected when the average luminosities of each of the miniature images have between them a difference less than a doublon threshold value, said duplicate threshold value is between 0 and 0.4, preferably between 0.05 and 0.2, and preferentially equal to 0.1.

This makes it possible to value the photos taken several times indicating the importance of the photo.

Advantageously, at least one duplicate is detected when the percentages of pixels belonging to a contour of said two miniature images mutually have a difference smaller than a threshold value of details, said threshold value of details is between 0 and 0.12, preferably between 0.015 and 0.06, and preferably equal to 0.03.

This makes it possible to value the photos taken several times indicating the importance of the photo.

Advantageously, at least one duplicate is detected when the average luminosities of each of the miniature images have between them a difference less than a doubling threshold value.

This makes it possible to value the photos taken several times indicating the importance of the photo.

Advantageously, at least one duplicate is detected when the instants of the recording on the non-transitory memory of each of the miniature images have a temporal difference between them less than a threshold value of time, said threshold value of time is between 0 seconds and 30 seconds, preferably between 5 seconds and 20 seconds, and preferably equal to 10 seconds.

This makes it possible to value the photos taken several times indicating the importance of the photo.

Advantageously, at least one duplicate is detected when the instants of the recording on the non-transitory memory of each of the miniature images have a time difference between them less than a threshold value of time.

This makes it possible to value the photos taken several times indicating the importance of the photo.

Advantageously, the doublon threshold value is between 0 and 0.4, preferably between 0.05 and 0.2, and preferably equal to 0.1.

This makes it possible to value the photos taken several times indicating the importance of the photo.

Advantageously, the threshold value of details is between 0 and 0.12, preferably between 0.015 and 0.06, and preferably equal to 0.03.

This makes it possible to value the photos taken several times indicating the importance of the photo.

Advantageously, the overall evaluation score is incremented by a partial evaluation score attributed to each of the original images among the at least one plurality of original images presenting duplicates and whose overall evaluation score is the highest. among the duplicates at the time of the duplicate detection step, said partial evaluation score being between 0 and 0.1 and being multiplied by the number of duplicates detected.

This makes it possible to value the photos taken several times indicating the importance of the photo.

Advantageously, the overall evaluation score is incremented by a partial evaluation score attributed to each of the original images among the at least one plurality of original images presenting duplicates and whose overall evaluation score is the highest. among the duplicates at the time of the duplicate detection step.

This makes it possible to value the photos taken several times indicating the importance of the photo.

Advantageously, the partial evaluation score attributed to each of the original images, among the at least one plurality of original images presenting duplicates and whose overall evaluation score is the highest among the duplicates at the time of the step detection of duplicates, is between 0 and 0.1 and being multiplied by the number of duplicates detected.

This makes it possible to value the photos taken several times indicating the importance of the photo.

Advantageously, the at least one plurality of analysis steps of the at least one original image comprises at least one step of determining the overall evaluation score of the at least one original image, said overall score of evaluating the at least one original image being between 0 and 1.

This makes it possible to enhance the images from a plurality of tests.

Advantageously, the at least one plurality of analysis steps of the at least one original image comprises at least one step of determining the overall evaluation score of the at least one original image.

This makes it possible to enhance the images from a plurality of tests.

Advantageously, the overall evaluation score of the at least one original image is between 0 and 1.

This makes it possible to enhance the images from a plurality of tests.

It is specified that in the context of the present invention, the term "pixel" or its equivalents are defined as the smallest constituent element of a digital photograph, a digital image. A pixel may be characterized by a plurality of numerical values such as brightness, saturation, contrast, and color, for example.

The term "brightness" or its equivalents are defined as a numerical magnitude varying for example between 0 and 255 and representing the perception of the luminous aspect of the pixel studied: the darker the pixel, the lower its luminosity.

The term "saturation" or its equivalents is defined as a numerical evaluation of the intensity of the coloration of a color, this value varying for example between 0 and 255.

It is specified that in the context of the present invention, the term "matrix" or its equivalents is defined as a table of data. A matrix has a size and a dimension, for example a matrix with m rows and n columns is called "size (m, n)", and a matrix with two inputs is called dimension 2, mathematically a matrix can be of dimension any. In the case where m is equal to n, we then speak of a square-shaped matrix.

Thus, the term "matrix of pixels" or its equivalents is defined by an array comprising a plurality of pixels so as to completely fill the said array. An image or digital photo can then be described as a matrix of pixels.

As a result, the term "central pixel area" of the pixel matrix or its equivalents is defined as a portion of pixels present in the central part of the matrix, that is to say distinct from the pixels forming the edges of the matrix. of pixels, and the term "outer pixel area" of the pixel matrix or its equivalents is defined as a portion of pixels present in the periphery of the pixel array, preferably surrounding the central pixel area.

The term "group of pixels" or its equivalents is defined as a set of pixels sharing a common color, or a brightness or a common saturation.

The term "contour" or its equivalents is defined as a plurality of pixels delimiting at least one group of pixels.

The term "sharpness" or its equivalents is defined as an increased perception of the details of a digital image, a sharp image is defined as the opposite of a fuzzy image.

It is pointed out that in the context of the present invention, the term "miniature" or its equivalents has a reduced image size compared to an original image, for example, a miniature pixel matrix is a smaller pixel matrix that an original pixel array, more particularly, a miniature pixel array comprises only a portion of a pixel array of an original image, that is, a thumbnail image is a portion of an original image.

It is specified that in the context of the present invention, the term "duplication" or its equivalents have the definition of making an integral copy of a matrix of pixels of an original image so as to form an identical matrix of pixels to the pixel matrix of the original image, said pixel array thus forming a duplicate image.

The present invention relates to a method of automatically selecting an original image (10) from a plurality of original images based on a global evaluation score assigned to said original image (10). This overall evaluation score is attributed to the original image (10) during the process according to the present invention. One of the objectives of the present invention is to allow an automatic selection of an original image (10) based on its visual quality, quality which is evaluated through a series of analysis steps and which is expressed through a overall evaluation score in the form of a numerical value.

Thus, when the invention is in operation on a mobile device, for example, a plurality of original images are loaded into the invention, then one after the other, they are analyzed and assigned a global evaluation score . Based on said overall evaluation score of each original image, various actions may be contemplated depending on the preferences of the user of the present invention.

Advantageously, when an original image (10) has its note set to zero, it is automatically excluded from the rest of the analysis process and preferably can not be selected.

Figure 1 depicts a preferred embodiment of the present invention. FIG. 1 illustrates the at least one plurality of analysis steps (100, 200, 300, 400, 500, 600, 700, 800) of an original image (10) among a plurality of original images allowing the automatically selecting an original image (10) from a plurality of original images based on its photographic quality and / or its interest to the user. The at least one plurality of analysis steps (100, 200, 300, 400, 500, 600, 700, 800) advantageously comprises a step of analyzing the resolution (100) of the original image (10) . This step eliminates the original images whose resolution, that is to say the total number of pixels composing the matrix of pixels of the original image (10), would be less than a resolution threshold value.

Advantageously, this resolution threshold value can be set by the user and / or preset. This threshold value of resolution is between 250,000 and 4,000,000, advantageously between 500,000 and 2,000,000, and preferably equal to 1,000,000.

This step of analyzing the resolution (100) of the original image (10) sets the overall evaluation score of the original image (10) to zero (120) if its resolution, that is to say its number of pixels is less than the threshold value of resolution. Therefore, the original image (10) considered is excluded from the rest of the analysis steps, and the present invention proceeds to the next original image (10).

According to a preferred embodiment, the pixel array of the original image (10) is defined by at least one central pixel area (1) and at least one outer pixel area surrounding said central pixel area (1).

According to a preferred embodiment, the at least one plurality of analysis steps (100, 200, 300, 400, 500, 600, 700, 800) comprises a step of generating (200) at least one image miniature lens (11) being formed by a miniature pixel array constituted by only a portion of said pixel array of the original image (10). Otherwise formulated, the thumbnail image (11) is a smaller size image than the original image (10) and thus represents only a portion of the original image (10).

Advantageously, the miniature image (11) is formed by cutting the original image (10) so as to retain only a portion of it.

In order to form this thumbnail image (11), several methods can be used.

Preferably, illustrated in FIG. 2, the step of generating (200) the miniature image (11) comprises a step of duplicating (210) the original image (10) so as to create a duplicated image (20). ). This duplicated image (20) is a copy, preferably perfect, of the original image (10), i.e. the pixel array of the duplicated image (20) is identical to the pixel matrix of the original image (10).

According to one embodiment, the duplicated image (20) can be a compressed version of the original image (10) so as to limit the memory space occupied by this duplicate image (20).

Advantageously, the generation step (200) also comprises a step of deleting (220, 230) a plurality of pixels from the matrix of pixels of the duplicated image (20) so as to form the at least one image miniature (11).

According to one embodiment, the deletion step (220, 230) is configured so that at least a portion of the central pixel area of the pixel array of the original image (10) is retained in the matrix of miniature pixels of the thumbnail image (11). Indeed, the central area of a photograph is usually the area of primary interest because what the user wants to photograph is usually in the center of the photograph. Advantageously, the present invention focuses its analyzes on the area of the original image (10) comprising the most interest for the user by keeping this area to form the miniature image (11).

According to another embodiment, the user can select the portion of the original image (10) that he wishes to keep in the miniature image (11) thus defining an area of interest. This option is useful in the case where the original image (10) is of interest not in its center but on the sides, this may be the case of a landscape picture with a person in the foreground shifted on the edge of the original image (10). In this type of situation the user can, if he wishes, himself define the area of interest to keep for the generation of the thumbnail image (11).

According to yet another embodiment, a step of analyzing the original image (10) can be performed before the step of generating (200) the thumbnail image (11) so as to locate in the original image (10) an area of interest, that is to say having remarkable characteristics such as the presence of a face, or particular colorimetric details, so that this area of interest is preserved in the miniature image (11).

Advantageously, the deletion step (220, 230) is configured to remove a periphery of pixels surrounding the central pixel area (1) of the original image (10) and / or the area of interest.

For example, this deletion of pixels can be achieved by removing bands of pixels (5) on either side of the central zone (1) and / or the area of interest.

According to one conceivable embodiment, it is at least part of the outer pixel area that is retained in the thumbnail image (11) during the blanking step (220, 230).

According to one embodiment, the deletion step (220, 230) is configured such that the thumbnail image (11) is formed by a square pixel array. Advantageously, the size of the miniature pixel matrix is between 75 pixels by 75 pixels and 1200 pixels by 1200 pixels, preferably between 150 pixels by 150 pixels and 600 pixels by 600 pixels, and preferably equal to 300 pixels by 300 pixels. pixels. Surprisingly, it has appeared that such a small matrix is sufficient to appropriately select the best original images from a plurality of original images. The use of a square pixel matrix has many algorithmic advantages that allow a reduction in the system resources required to execute the present invention. Indeed, the manipulation of a square matrix simplifies the methods of computation and analysis of the components of said square matrix.

In addition, the use of a thumbnail image (11) allows many advantages in terms of computation time but also memory resources since a miniature image comprises fewer pixels than an original image, it is therefore more easily and quickly manipulable. Moreover, because of a possible plurality of sources of the different original images, their size may vary, making miniature images of known and fixed size then allows optimization of the automatic selection by not having to adapt to the image. and thus not having to consume additional system resources.

According to one embodiment of the present invention, it is the image that adapts to the system and not the other way around.

This thumbnail image (11) is then used in a plurality of steps to be analyzed and to allow the evolution of the overall evaluation score of the original image (10) during the selection process.

According to one embodiment, the at least one plurality of analysis steps (100, 200, 300, 400, 500, 600, 700, 800) comprises a step of evaluating the average brightness of the miniature image (11). This step may for example be performed by averaging the brightness of each pixel of the miniature pixel array.

Advantageously, in the case where the value of the average brightness of the miniature image is greater than a maximum brightness threshold value, the overall evaluation score is set to zero (320), and the original image (10) is then excluded from the rest of the selection process. This threshold value of maximum brightness is between 25% and 95%, advantageously between 45% and 85%, and preferably equal to 70%. This makes it possible to eliminate from the selection the original overexposed images.

Advantageously, in the case where the value of the average brightness of the miniature image is less than a minimum brightness threshold value, the overall evaluation score is set to zero (320), and the original image (10) is then excluded from the rest of the selection process. This threshold value of minimum brightness is between 5% and 55%, advantageously between 15% and 35%, and preferably equal to 20%. This makes it possible to eliminate from the selection the original underexposed images.

According to one embodiment, the at least one plurality of analysis steps (100, 200, 300, 400, 500, 600, 700, 800) comprises a sharpness evaluation step (400) of the miniature image (11). This step then eliminates original images of which a portion, the thumbnail image (11), is estimated too fuzzy, that is to say not clear enough.

Advantageously, this step relies on an edge detection technique well known to those skilled in the art. However, this edge detection is performed only on the thumbnail image (11) so as to save energy and time during the execution of this step.

During this step of evaluating the sharpness (400) of the miniature image (11), the number of pixels belonging to a contour delimiting two distinct groups of pixels is evaluated. Then this number is reduced to a percentage with respect to the number of pixels of the miniature pixel array. This percentage then allows the incrementation of the overall evaluation score.

In the case where this percentage of pixels is less than a minimum contour threshold value, the overall evaluation score of the original image (10) is set to zero (420), and the original image (10) is then excluded from the rest of the selection process. This minimum threshold threshold value is between 0.5% and 4.5%, advantageously between 1.5% and 3.5%, and preferably equal to 2.5%. This then makes it possible to eliminate from the selection the fuzzy original images.

In the case where this percentage is greater than this minimum contour threshold value, the overall evaluation score of the original image is incremented by a partial evaluation score N1. This evaluation partial score N1 corresponds to the percentage of pixels belonging to an outline in the miniature image (11).

According to a preferred embodiment, the partial note N1 has no limit value.

According to a particular embodiment, the partial score N1 is limited to a maximum value of 0.25 in order to maintain a uniform evaluation among the different steps of the present invention.

Thus according to a particular embodiment, the partial note N1 is between 0 and 1, and advantageously between 0.1 and 0.5.

According to one embodiment, a coefficient k1 can be assigned to this analysis step so as to weight the participation of this partial evaluation score in the overall evaluation score.

Advantageously, k1 is equal to 2. This makes it possible to reinforce the overall evaluation score of the original images comprising a high percentage of pixels belonging to a contour so as to reinforce the overall evaluation score of the original images presenting numerous visual details, and therefore a visual richness.

According to a preferred embodiment, the at least one plurality of analysis steps (100, 200, 300, 400, 500, 600, 700, 800) comprises an additional analysis step (500) of at least a plurality of pixels composing the miniature pixel array of the thumbnail image (11). This additional analysis step (500) is configured to evaluate visual parameters, preferably, of each pixel constituting the miniature pixel matrix. These visual parameters can for example be the color, the brightness, the contrast, the saturation or any other data that can characterize a pixel.

Advantageously, the present invention preferably uses the brightness and saturation data of each of the pixels of the miniature pixel array.

According to a preferred embodiment of the present invention, said additional analysis step (500) is configured to evaluate the brightness and / or saturation of each pixel composing the miniature pixel array.

Advantageously, the percentage of pixels of the miniature pixel matrix whose brightness is greater than a saturated brightness threshold value and / or whose saturation is greater than a saturation threshold value, is evaluated so as to increment the overall score of evaluating the original image (10) by a partial evaluation score N2 corresponding to said percentage of pixels. Preferably, said partial evaluation score N2 is between 0 and 1, and preferably between 0.45 and 0.65.

According to one embodiment, a coefficient k2 can be assigned to this analysis step so as to weight the participation of this partial evaluation score N2 in the overall evaluation score.

According to one embodiment, the evaluation partial score N2 is limited to a maximum value of 0.5 in order to maintain a uniform evaluation among the different steps of the present invention.

According to a particular embodiment, said threshold value of saturated brightness is between 25% and 75%, preferably between 40% and 60% and preferably equal to 50%.

According to a particular embodiment, said saturation threshold value is between 25% and 75%, preferably between 40% and 60% and preferably equal to 50%.

This additional analysis step (500) thus makes it possible to identify the percentage of so-called saturated pixels and thus to increment the overall assessment score of the original image (10) by a value reflecting the richness of the visual dynamics. from the original image (10).

According to one embodiment, the at least one plurality of analysis steps (100, 200, 300, 400, 500, 600, 700, 800) comprises a face detection step (600). This face detection step is configured to evaluate the number of faces present in the original image (10). For this optional step, the original image (10) is analyzed.

Advantageously, for each detected face, the overall evaluation score of the original image (10) is incremented by a partial evaluation score N3. Said evaluation partial score N3 is a function of the number of detected faces, preferably a numerical value is assigned to each face. This numerical value is preferably between 0.05 and 0.5, advantageously between 0.1 and 0.25, and preferably equal to 0.15.

According to one embodiment, a coefficient k3 can be assigned to this analysis step so as to weight the participation of this partial evaluation score in the overall evaluation score.

According to one embodiment, the partial evaluation score N3 is limited to a maximum value of 0.5 in order to maintain a uniform evaluation among the different steps of the present invention.

Advantageously, a step of detecting faces with a smile can be implemented. In this step, the partial evaluation score N3 is different in the case where the face has a smile or not. The partial evaluation score N3 may then be between 0.1 and 0.5, advantageously between 0.15 and 0.25, and preferably equal to 0.2.

According to one embodiment, the at least one plurality of analysis steps (100, 200, 300, 400, 500, 600, 700, 800) comprises a duplicate detection step (700) among the at least one a plurality of original images, ie the original image detection having very similar similarities, or even being identical to each other.

According to one embodiment, this duplicate detection step (700) is configured to detect a duplicate when the average brightnesses of two thumbnails have a difference between them less than a duplicate value, and / or when the times of the recording on the non-transitory memory of said two original images has a temporal difference between them less than a threshold value of time, and / or when the percentages of pixels belonging to a contour of the two miniature images have a difference between them less than one threshold value of details. According to one embodiment, the instant of recording on the non-transitory memory of said two original images may correspond to the moment of creation of said original images and not of the moment of recording.

Advantageously, the doublon threshold value is between 0 and 0.4, preferably between 0.05 and 0.2, and preferably equal to 0.1.

Advantageously, the threshold value of details is between 0 and 0.12, preferably between 0.015 and 0.06, and preferably equal to 0.03.

Advantageously, the threshold value of time is between 0 seconds and 30 seconds, preferably between 5 seconds and 20 seconds, and preferably equal to 10 seconds.

This step benefits from the result of the step of evaluating the average brightness (300) of the miniature images, which allows a very considerable saving of processing time. The use of a temporal parameter makes it possible to increase the relevance of this duplicate detection step (700). Similarly, the use of the contour pixels makes it possible to increase the relevance of this duplicate detection step (700).

According to one embodiment, the duplicate detection step (700) comprises a step of selecting an original image (10) whose overall score is the highest among the duplicates detected for this original image (10) at the time performing this duplicate detection step (700). Among a plurality of duplicates, the present invention selects the one whose overall evaluation score is the highest of all in order to increment it with a partial evaluation score N4 as a function of the number of duplicates detected.

Advantageously, the partial evaluation score N4 is between 0 and 0.1 multiplied by the number of duplicates detected.

According to one embodiment, a coefficient k4 can be assigned to this analysis step so as to weight the participation of this partial evaluation score in the overall evaluation score.

According to one embodiment, the evaluation partial score N4 is limited to a maximum value of 0.5 in order to maintain a uniform evaluation among the various steps of the present invention.

This step allows highlighting original images representing important scenes for the user since taken several times in photo.

This duplicate detection step (700) has a significant advantage in non-transitory memory management. Indeed, from this duplicate analysis, it is possible to remove the other duplicates so as to release memory space according to the parameters of the present invention.

According to one embodiment, the at least one plurality of analysis steps (100, 200, 300, 400, 500, 600, 700, 800) comprises a step of determining (800) the overall evaluation score from the original image (10). This overall evaluation score is calculated from values of the partial scores N1, N2, N3 and N4 and their respective coefficients k1, k2, k3 and k4.

Advantageously, the overall evaluation score is between 0 and 1. According to a preferred embodiment, the overall evaluation score is calculated at the end of each of the plurality of analysis steps (100, 200, 300, 400, 500, 600, 700, 800).

According to one embodiment, following this at least a plurality of analysis steps (100, 200, 300, 400, 500, 600, 700, 800), a selection step (900) is configured to select the original images with the highest overall ratings. For this selection step (900) various parameters can be taken into account depending on the user and / or the parameters of the present invention.

Advantageously, all the different choices and decisions of the user may for example be recorded and / or prerecorded so that the invention executes the same choices as those of the user automatically.

For example, the user may decide to keep only a percentage of the plurality of original images, in which case the selection step (900) is configured to select the number of original images corresponding to that percentage according to a decreasing overall rating order rating. Similarly, the user can decide to select only a specific number of original images. The user may also decide to select only the original images whose overall score is greater than a threshold threshold value.

Advantageously, the threshold value of a score is between 0 and 1, preferably between 0.1 and 0.75, and preferably equal to 0.2. Conversely, the user can decide to select only the original images whose overall evaluation score is less than this threshold threshold value.

Finally, the user can decide to select only those images whose overall evaluation score is zero.

According to one embodiment, following this selection step, a management step of and / or selected original images (1000) is advantageously presented in FIG. 1. This step is configured to execute one or more actions according to the preferences of the user and / or parameters of the present invention, for example. In one possible case the selected image can be saved in a particular file, or in an album designed for this purpose. Duplicates of this image can be deleted, for example.

According to one embodiment, the selection can be made from the original images having overall evaluation scores very low or zero. In this case, these original images can be deleted because they are not interesting. This then allows a gain in large memory space by automatically and reliably removing original images of poor visual quality.

The present invention thus allows a significant gain in speed and system resource saving energy during the process, and allows a gain memory at the end of the process.

According to one embodiment, the present invention may be described as follows, a user of a smartphone-type device, for example, makes a plurality of photos with his smartphone, he also downloads and receives photos of his friends or family. . Thus, for example, during a wedding he will have a plurality of images that he has made himself, but also images made by other mobile devices of the smartphone type or even cameras. Therefore, its photo gallery will include a very high density of images but also a very large variety of format and / or image size. Indeed the resolution of the various photos may vary depending on the devices that captured them. From this multitude of photos that include both identical photos in terms of captured scenes, but also overexposed and / or underexposed images, blurred images, that is to say that are not clear, the user would spend a lot of time if he had to check one by one each photo.

Using the present invention, the user can automatically select the best photos from his gallery while doing something else with his smartphone. Indeed, the present invention then functions in the background because the system resources it needs are low.

Therefore, the user going to take the plane, for example, to return from said marriage puts his smartphone in airplane mode, that is to say it disconnects any connection to a communication network. While his flight is happening, the user can launch the automatic selection of the 100 best pictures of the wedding. Indeed, the present invention operating according to a local mode, it does not require connection to a communication network.

Before landing, the user has a file containing the 100 best photos of the wedding ready to be sent to a print shop as soon as the network is back to make a physical album for example. The user can then choose to delete duplicates or not, the photos whose overall rating is zero or less than a threshold value that it defines itself for example.

In order to personalize the automatic selection process, the user can adjust the various parameters of the various analysis steps described above in the description, that is to say all the threshold values for example.

The present invention relates to a method, a method, of automatically selecting an image from a plurality of images, said method being executed by a processor preferably present in a mobile device. Such a mobile device may include a smartphone, a digital tablet, a laptop, or any electronic equipment that can be mobile.

According to one embodiment, the present invention is also adapted to operate on a non-mobile computer or server type device.

The present invention is configured to be implemented by at least one processor in connection with at least one non-transient memory and a power source. Advantageously, the present invention is configured to limit the consumption of system resources during the selection process but also after the selection process. Indeed, the present invention allows an economy of the resources of the at least one processor by the use of an image of reduced size compared to an original image during the process. Similarly, according to one embodiment, it is the memory space of the non-transitory memory that can be saved when the user selects the option of deleting the images whose overall evaluation score would be lower than a chosen value. .

Thus the present invention allows an economy of the resources of the system both during the process but also at the end of it.

In parallel, the economy of the processor cycles also ensures a significant energy saving in the case of the use of the present invention on a smartphone type mobile device for example. At a time when everyone is running after an electrical outlet to recharge their phone, the present invention allows an optimization of the battery during the process and an optimization of the non-transitory memory at the end of the process if the user wishes.

According to a preferred embodiment, it is the original image that is configured for the automatic selection method in order to be executable on a mobile device for example, by consuming some system resources. The invention is not limited to the previously described embodiments and extends to all the embodiments covered by the claims.

Claims (16)

A method of automatically selecting at least one original digital image (10) from at least a plurality of digital original images, said at least one original image (10) being defined by a matrix of pixels, each pixel being defined by at least one numerical value of brightness and saturation, said method being configured to be executed by at least one processor included in at least one mobile device comprising at least one non-transient memory, at least one power source, and said method characterized in that it comprises at least: - A plurality of analysis steps (100, 200, 300, 400, 500, 600, 700, 800) of the at least one original image (10), said method being configured to assign a global evaluation score to the original image (10), the overall evaluation score being a numerical value incremented by at least a partial score evaluation score misting the original image (10) in at least one step of said plurality of analysis steps (100, 200, 300, 400, 500, 600, 700, 800); A step of generating (200) at least one miniature image (11) formed by a miniature pixel matrix constituted by a portion of said pixel matrix of the original image (10); A selection step (900) based on said overall score of the original image (10) and wherein at least one of said plurality of analysis steps (100, 200, 300, 400, 500, 600, 700, 800) is configured to use said thumbnail image (11). 2. Method according to the preceding claim wherein the step of generating (200) the at least one miniature image (11) of the at least one original image (10) comprises: - a step of duplication (210) of the at least one original image (10) so as to create at least one duplicated image (20), said at least one duplicated image (20) comprising a plurality of pixels forming a pixel array and - a deletion step (220) 230) of at least a plurality of pixels of the pixel array of the duplicate image (20) so as to form the at least one thumbnail image (11). 3. Method according to the preceding claim wherein the pixel array of the original image (10) is defined by at least one central pixel area and at least one outer pixel area surrounding said central pixel area and wherein the step of removing (220, 230) said at least one plurality of pixels is configured such that the thumbnail image (11) is formed by a square miniature pixel array. A method according to any one of the preceding claims wherein the size of the miniature pixel array is between 75 pixels by 75 pixels and 1200 pixels by 1200 pixels, preferably between 150 pixels by 150 pixels and 600 pixels by 600 pixels, and preferably equal to 300 pixels by 300 pixels. A method according to any one of the preceding claims wherein the pixel array of the original image (10) is defined by at least one central pixel area and at least one outer pixel area surrounding said central pixel area and wherein the step of generating (200) the thumbnail image (11) is configured such that at least a portion of the central pixel area of the original image (10) is retained in the miniature pixel array of the miniature image (11). The method according to any one of the preceding claims wherein the at least a plurality of analysis steps (100, 200, 300, 400, 500, 600, 700, 800) of the at least one original image. (10) comprises at least one step of analyzing the resolution (100) of the original image (10) comprising an evaluation of the number of pixels of the at least one original image (10), and wherein the overall rating of the original image (10) is set to zero when the number of pixels of the at least one original image (10) is less than a resolution threshold, said resolution threshold being between 250,000 pixels and 4,000,000 pixels, preferably between 500,000 and 2,000,000 pixels, and preferably equal to 1,000,000 pixels. The method according to any one of the preceding claims wherein the at least a plurality of analysis steps (100, 200, 300, 400, 500, 600, 700, 800) of the at least one original image. (10) comprises at least one step of evaluating a mean brightness (300) of the thumbnail image (11), said average brightness of the thumbnail image (11) being calculated by averaging the brightness of each pixel of the miniature pixel array, and wherein the overall evaluation score of the original image (10) is set to zero when the average brightness of the thumbnail image (11) is greater than a maximum brightness threshold value and / or when the average brightness of the thumbnail image (11) is less than a minimum brightness threshold value. The method according to any one of the preceding claims wherein the at least a plurality of analysis steps (100, 200, 300, 400, 500, 600, 700, 800) of the at least one original image. (10) comprises at least one step of evaluating the sharpness (400) of the thumbnail image (11). 9. Method according to the preceding claim wherein the at least one sharpness evaluation step (400) is based on an edge detection technique, a contour being defined as a set of pixels delimiting two continuous zones of pixels, and wherein the percentage of pixels, relative to the miniature pixel array, belonging to a contour is evaluated, and wherein the overall evaluation score of the original image (10) is set to zero when said percentage of pixels is less than a minimum contour threshold value, and wherein the overall evaluation score is incremented by a partial evaluation score of the sharpness evaluation step (400) attributed to the original image (10). ), said partial evaluation score of the sharpness evaluation step (400) being a function of said percentage of pixels. The method of any one of the preceding claims wherein the at least a plurality of analysis steps (100, 200, 300, 400, 500, 600, 700, 800) of the at least one original image. (10) comprises at least one additional analysis step (500) of at least a plurality of pixels composing the miniature pixel array of the thumbnail image (11), said additional scanning step (500) being configured to evaluating the brightness and / or saturation of each pixel constituting the miniature pixel matrix, and in which the percentage of pixels of the miniature pixel matrix, the brightness of which is greater than a saturated brightness threshold value and / or whose saturation is greater than a saturation threshold value, is evaluated so as to increment the overall evaluation score of the original image (10) by a partial evaluation score corresponding to the said percentage of pixels. The method according to any of the preceding claims wherein the at least a plurality of analysis steps (100, 200, 300, 400, 500, 600, 700, 800) of the at least one original image. (10) comprises at least one face detection step (600) present in the original image (120), and wherein for each detected face, the overall evaluation score is incremented by a partial evaluation score assigned to the original image (10), said partial evaluation score being a function of the number of faces detected. The method according to any one of the preceding claims wherein the at least a plurality of analysis steps (100, 200, 300, 400, 500, 600, 700, 800) of the at least one original image. (10) comprises at least one duplicate detection step (700) among the at least one plurality of original images. The method according to the preceding claim wherein the plurality of original images are recorded on said non-transitory memory of the mobile device and wherein the duplicate detection step (700) is configured to detect a duplicate when the luminosities means of two miniature images have a difference between them less than a doublet threshold value, and when the instants of the recording on the non-transitory memory of said two original images (10) have a time difference between them less than a threshold value of temporality, said threshold value of time being between 0 seconds and 30 seconds, preferably between 5 seconds and 20 seconds, and preferably equal to 10 seconds, and when the percentages of pixels belonging to a contour of said two miniature images have a difference between them less than a threshold value of details and in wherein the duplicate detection step (700) includes a step of selecting an original image (10) whose overall evaluation score is the highest of the duplicates detected for that original image (10) at the time of the duplicate detection step (700), the overall assessment score of the selected original image (10) is incremented by a partial evaluation score. The method according to any of the claims wherein the at least a plurality of analysis steps (100, 200, 300, 400, 500, 600, 700, 800) of the at least one original image ( 10) comprises at least one step of determining (800) the overall evaluation score of the at least one original image (10). A computer program product stored in a non-transitory memory and having instructions executable by at least one processor, configured to execute the method of any one of the preceding claims. 16.Mobile terminal comprising an image pickup sensor (10), typically a camera, at least one processor, at least one power source and at least one non-transitory memory configured to save the images taken by the sensor and at least one memory storing a computer program product having instructions executable by said at least one processor, said instructions being configured to execute the method of any one of claims 1 to 14.