KR101459782B1 - A system for enhancing a night time image for a vehicle camera - Google Patents

Abstract

This document is to improve nighttime image of vehicle camera which improves the details of image in dark region by judging noise area through difference of pixel value between night and night image of car camera and applying image improvement algorithm only to non- ≪ / RTI > According to the method of improving the night image of the vehicle camera, the object identification ability of the camera is improved in the dark area.

Night vision, retinex, noise area

Description

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a nighttime image enhancement system for a car camera,

This document relates to a vehicle camera, and more specifically, to a method for improving a night image of a car camera.

In recent automobile technology, a compact camera is installed on the front and rear sides including the left and right sides for safety during operation, and a system for confirming the left and right side and front and rear sides through the display of the driver's seat panel has been applied .

As the use of cameras in automobiles becomes full-scale, the image quality of output images is becoming an important measure for consumers to buy cameras. In the case of today's automotive cameras, problems with camera modules such as data compression techniques, power consumption, circuit problems such as limitations of built-in memory, difficulty of optical zoom, use of low-band optical filters, low level color reproducibility, The image reproduction performance is very low compared to the dedicated digital camera.

Particularly, it is not easy to recognize an object with respect to a night image, and image quality is further deteriorated due to influence of noise or the like. Therefore, in order to secure the safety of the driver when the vehicle is running, an improvement algorithm for the night image is essential.

This document aims to improve the object identification ability of the camera in the dark region by improving the night image of the vehicle camera.

A nighttime image enhancement system for a vehicle as a means for solving the above problems is characterized by including a brightness value converter for converting RGB values of a night image into brightness values, And a noise region determination unit for determining a noise region and a non-noise region by comparing pixel values with pixels.

The nighttime image enhancement system of the vehicle camera includes an image processor for processing the nighttime image by applying an image enhancement algorithm only to pixels included in a region determined as a noise-free region through the noise region determination unit.

The noise region determination unit may determine whether the pixel is noise by applying a noise identification mask to the pixel.

Wherein the noise region determination unit determines that the center pixel is in a noise region when at least one of a plurality of difference values calculated by comparing a center pixel of the noise identification mask with a pixel value of a surrounding pixel exceeds a predetermined threshold value .

The noise region determination unit may determine that the center pixel corresponds to the noise region when the difference value corresponding to the third or fourth is greater than the predetermined threshold value .

The noise region of the night image may be image-processed using an initial value.

As the image enhancement algorithm, Retinex algorithm can be applied.

According to the present invention, by applying the Retinex algorithm in the dark area to improve the night image of the car camera, the detail information of the image is improved, thereby securing the safety at night.

In addition, by applying the image enhancement algorithm by separating the noise region and the non-noise region, the image enhancement algorithm for the noise region can be avoided, thereby enhancing the image enhancement effect.

In addition, there is an effect that a noise region can be determined and used by a simple algorithm using a noise identification mask.

Hereinafter, embodiments of the present invention will be described in more detail with reference to the accompanying drawings.

1 shows a block diagram of a night vision image enhancement system for a vehicle camera according to an embodiment of the present invention.

The nighttime image enhancement system for a vehicle according to the present embodiment converts RGB values of a video image, in particular, a night image from a video input unit 10 and an image input unit 10, And a brightness value converting unit 20 for generating an image to be displayed. Equation 1 below shows an example of a method of converting the RGB values of the night image input to the brightness value converter 20 into brightness values according to an embodiment of the present invention.

In Equation (1), Y represents a brightness value. As shown in Equation (1), since the G value among the RGB values has the greatest influence on the brightness, it is preferable to give the most weight to the G value. It is to be understood that the concrete numerical values of the equation (1) can be changed at any time and are not limited to those shown in the equation (1).

The nighttime image enhancement system of the vehicle camera according to the present embodiment includes a noise region determination unit 30 that determines a noise region using an image represented by a brightness value generated by the brightness value conversion unit 20.

The noise region determination unit 30 according to the present embodiment compares the pixel value of each pixel of the image represented by the brightness value with the pixel value of surrounding pixels and if the difference value exceeds the predetermined threshold value, Area. This is because when a pixel has a large difference between neighboring pixels and a pixel value, the probability that the pixel is determined to be noise is high.

The nighttime image enhancement system of the vehicle camera according to the present embodiment is a system for improving the night image of a vehicle camera by dividing a region judged to be a noise region and a region judged to be a noise-free region by a noise region determination unit 30, And an image processing unit 40 for processing an image by applying the image processing unit.

That is, in the image processing unit 40 according to the present embodiment, general image processing is performed on an area determined as a noise area, and the image is processed by applying an image enhancement algorithm only on an area determined as a noise-free area.

In addition, the nighttime image enhancement system for a vehicle camera according to the present embodiment further includes an image output unit 50 for acquiring an image processed by the image processing unit 40 and outputting the processed image.

FIG. 2 illustrates an example of a noise identification mask applied to an image represented by a brightness value converted by a noise region determination unit according to an embodiment of the present invention.

In nighttime images, that is, dark images, there is a lot of noise in the car camera images. Therefore, according to the present embodiment, the brightness value converting unit 20 converts the RGB values of the image into brightness values, and determines a noise region using the noise identification mask on the night image generated based on the brightness values.

In the noise region determination unit 30 according to the present embodiment, the noise identification mask may be applied to the image to determine the noise region and the noise-free region. The noise identification mask is used to define a certain pixel region of an image and define a specific pixel and surrounding pixels of the specific pixel.

According to this embodiment, a noise identification mask is applied to each pixel of an image represented by a brightness value, and each pixel is applied so as to be a center pixel of a noise identification mask. Here, the brightness value can be obtained by the method described above using Equation (1). Then, the center pixel of the noise identification mask is compared with the pixel value of the neighboring pixel to determine the noise-containing pixel.

A 3x3 noise identification mask is shown in Fig. According to this, eight peripheral pixels are defined for the center pixel. According to this embodiment, it is determined whether at least one of the eight difference values calculated by comparing the pixel value of the central pixel with the pixel value of the eight surrounding pixels exceeds a predetermined threshold value, Or a non-noise region.

In the noise identification mask shown in Figure 2 x _{i, j} is the current (i, j) corresponding to the location represents a pixel value for the pixel, w0, w1, w2, w3 , w4, w5, w6, w7 is (i , J) represents the pixel value of the surrounding pixels of the pixel corresponding to the position.

That is, the difference value between each neighboring pixel is calculated by sequentially comparing x _{i, j} with w 0, w 1, w 2, w 3, w 4, w 5, w 6, w 7. Then, it is possible to determine whether the pixel corresponding to the position (i, j) corresponds to a noise region or a non-noise region by comparing the difference values with a predetermined threshold value.

In particular, all of the difference values generated as a result of comparison with neighboring pixels can be compared with predetermined threshold values, but only a part of the plurality of difference values may be compared with a predetermined threshold value.

3 illustrates a method of determining a noise region using a noise identification mask according to an embodiment of the present invention.

For example, as shown in FIG. 3, in step S30, a plurality of difference values generated as a result of comparison with neighboring pixels are sorted in ascending order or descending order. Then, in step S32 and step S33, And can be compared with a preset threshold value. At this time, the sorted difference values may be sequentially numbered and used as in step S31.

If the fourth difference value exceeds the preset threshold value as in step S32, it is determined in step S34 that the pixel is a noise area. If it is determined in step S33 that the pixel is not a noise area, .

Here, it is needless to say that whether to compare the difference value corresponding to a certain size with a preset threshold value is not limited to the fourth, but may be variously determined, for example, third, fifth, and so on.

4 illustrates a method of applying an image enhancement algorithm by distinguishing a noise region from a noise region according to an embodiment of the present invention.

3, the image enhancement algorithm is applied to the non-noise region, and the noise region is subjected to the image processing using the initial value as it is without applying the image enhancement algorithm. The reason for this is that applying the image enhancement algorithm to the noise region will improve the noise and lower the image enhancement performance.

Here, as the image enhancement algorithm, the Retinex algorithm that compensates the image using only the reflection image of the image can be used. Here, according to the Retinex algorithm, the roughness component is obtained before the reflection component is obtained. The roughness component is considered as a low-frequency component under the assumption that the roughness component changes slowly, and can be obtained through a low pass filter (LPS). Then, the illumination component is separated by taking logs on both sides in the expression model represented by the illuminance * reflection component. In this way, the difference between the log signal of the original image and the illuminance log signal is obtained, and the reflection component can be extracted. Then, the image can be corrected using the extracted reflection component.

Accordingly, in the present invention, the noise region is identified using the noise identification mask with respect to the image input in Step S100, that is, the night image of the vehicle camera, and in Step S110, the noise region is advanced to Step S120, If the image is a noise-free region in step S130, the process proceeds to step S140 and image processing using the Retinex algorithm is performed. Then, in step S150, the images of the noise region and the noise-free region, which are processed individually, are combined and an image is output. That is, according to the present embodiment, by applying the Retinex algorithm only to the non-noise region, the detail of the image is improved, thereby securing the safety of the driver at night.

FIG. 5 shows an example of a result of applying the night image improving method according to an embodiment of the present invention.

5 (a) and 5 (b), the left side shows the image before applying the night image improving algorithm, and the right side shows the applied image. The image on the left is dark and it is very difficult to judge the object accurately, whereas the image on the right appears to be able to identify the objects in the image to some extent. As can be seen from FIG. 5, it can be seen that a very high performance image improvement is achieved by applying the image enhancement algorithm by distinguishing the noise region from the non-noise region.

It will be apparent to those skilled in the art that various modifications and variations can be made in the present invention without departing from the spirit and scope of the invention as defined by the appended claims. You will understand.

BRIEF DESCRIPTION OF THE DRAWINGS Fig. 1 shows a block diagram for night image enhancement of a car camera according to an embodiment of the invention.

FIG. 2 illustrates an example of a noise identification mask applied to an image represented by a brightness value converted by a noise region determination unit according to an embodiment of the present invention.

3 illustrates a method of determining a noise region using a noise identification mask according to an embodiment of the present invention.

FIG. 4 illustrates a method of applying a Retinex algorithm by separating a noise region from a noise region according to an embodiment of the present invention.

FIG. 5 shows an example of a result of applying the night image improving method according to an embodiment of the present invention.

Description of the Related Art

10: image input unit 20: brightness value conversion unit

30: noise region determining unit 40: image processing unit

50: Video output section

Claims (6)

A brightness value converter for converting the RGB values of the night image into brightness values; A noise region determination unit for determining a noise region and a non-noise region by comparing pixel values of pixels and surrounding pixels of the pixel in the night image represented by the brightness values; And An image enhancement algorithm is applied to the night image, and the image enhancement algorithm is applied only to the noise-free area of the night image to process the night image, And a nighttime image enhancement system for a vehicle camera. The method according to claim 1, The noise region determination unit may determine whether the pixel is noise by comparing the pixel value of the central pixel with the pixel value of the surrounding pixel by applying a noise identification mask defining the center pixel and the surrounding pixels to the pixel, And the nighttime image enhancement system of the camera for a vehicle. The method of claim 2, Wherein the noise region determination unit determines that the center pixel is in a noise region when at least one of a plurality of difference values calculated by comparing a center pixel of the noise identification mask with a pixel value of a surrounding pixel exceeds a predetermined threshold value The nighttime image enhancement system of the vehicle camera. The method of claim 3, The noise region determination unit may sort the plurality of difference values in order of magnitude and if the difference value corresponding to the third or fourth is greater than the predetermined threshold value, the center pixel is determined to correspond to the noise region And the nighttime image enhancement system for a car camera. The method according to claim 1, Wherein the image processing unit comprises: Wherein the nighttime image processing unit processes the nighttime image without applying the image enhancement algorithm to the noise region of the nighttime image. The method according to claim 1, Wherein the image enhancement algorithm is a Retinex algorithm.

Images