KR20160005320A - Recording medium, portable terminal and method for recognizing characters based on determining presence of user's hand tremor or intentional motion - Google Patents

Abstract

According to an aspect of the present invention, a method for recognizing a character comprises the following steps: determining a sample block in a previous frame of a preview image; searching a matching block matched with the sample block in a current frame; finding a first block distance between the sample block and the matching block; and performing a character recognizing operation when the first block distance is less than or equal to a first threshold value.

Description

TECHNICAL FIELD [0001] The present invention relates to a character recognition method, a recording medium, and a portable terminal based on judgment of presence or absence of a user's hand motion or intentional motion, a recording medium,

The present invention relates to an apparatus and method for detecting a motion of a subject in input images, and more particularly, to a method for determining whether a user's hand motion or intentional motion exists during image recognition through photographing.

The most commonly used methods for detecting motion from moving images are block matching, optical flow, and the like. In general, block matching through the sampling step is fast and the optical flow method through the optimization step is slower than the block matching but shows a more accurate performance. As a high-speed motion detection method, UMHexagonS (Unsymmetric-cross Multi-Hexagon-grid Search) method, which is known to be the best in terms of experimental results, is widely used as a block matching method. For example, the UMHexagonS method is described in detail in Zhibo Chen, et al., "Fast integer-pel and fractional-pel motion estimation for H.264 / AVC", Journal of Visual Communication & Image Representation .

The UMHexagonS search method samples the search area according to each step. In this sampling stage, we first search the global scope and narrow down the search locally. However, there is a need for a method for determining presence or absence of motion suitable for the use environment for an application such as a mobile camera based interest area character recognition, area segmentation and analysis. For this purpose, a precise motion detection technique in the region of interest is required, which is faster than in the prior art. As a method for improving the calculation speed, a method of reducing the size of the block to be matched can be considered.

FIG. 1 is a view for explaining differences in motion estimation according to the size of a block to be matched. In FIG. 1, the matching target blocks 10 and 20 before the motion are represented by dotted lines, and the matching target blocks 12 and 22 after the motion are indicated by solid lines. As shown in the figure, in the case of recognizing the characters on the ground, when the large matching target blocks 20 and 22 are used, the matching target blocks 20 and 22 are likely to include the edge on the ground, The accuracy is high but the detection speed is slow. On the contrary, when the small matching target blocks 10 and 12 are used, the possibility of the matching target blocks 10 and 12 not including the edges on the ground is high, so that the motion detecting speed is fast but the detection precision is low. In addition, the hexagonal search method or diamond search method has low detection accuracy.

Accordingly, there is a need for a method and apparatus for determining the presence or absence of a user's hand motion or intentional motion during image recognition through photographing that can guarantee a high detection speed and high precision required in a portable terminal such as a mobile phone.

According to an aspect of the present invention, a character recognition method includes: detecting a sample block in a previous frame of a preview image; Searching a matching block matching the sample block in a current frame; Obtaining a first block distance between the sample block and the matching block; And performing a character recognition operation when the first block distance is equal to or less than the first threshold value.

According to another aspect of the present invention, there is provided a method of determining the presence or absence of a user's hand motion or intentional motion during image recognition through photographing, includes: detecting image sample blocks of a previous frame using an edge detection method; Obtaining a block distance between the detected image sample block and an estimated image sample block of a current frame; A first comparing step of comparing the obtained block distance with a first threshold value; And a second comparing step of comparing the obtained block distance and a second threshold value when the obtained block distance is smaller than the first threshold value in the first comparing step.

Preferably, in the first comparison step, it is determined that there is an intentional movement of the user when the obtained block distance is greater than the first threshold value.

Preferably, in the second comparison step, it is determined that there is an intentional movement of the user when the obtained block distance is greater than the second threshold value, and when the obtained block distance is less than the second threshold value, .

Advantageously, the first threshold is a value that is greater than a tolerance allowed for motion between the detected image sample block and the estimated image sample block of the current frame, Is the allowable shake tolerance for motion between the estimated image sample blocks of the current frame.

Preferably, in the detecting of the image sample block, image sample blocks are detected using an edge detection method for a plurality of blocks obtained by dividing a predetermined area of the previous frame.

Preferably, in the step of obtaining the block distance, a block distance is estimated through a comparison matching of the image sample blocks in the previous frame and the neighboring blocks in the current frame.

Preferably, the method further comprises calculating an intermediate value of the block distances obtained in the step of obtaining the block distance, and when the determined block distance in the first comparison step is smaller than the first threshold value, And compares the intermediate value with the second threshold value.

Preferably, in the detecting step of the image sample block, the image sample blocks including the edge component of the plurality of blocks are detected, and a predetermined number of images in order of increasing edge components in the detected image sample blocks Select sample blocks.

Preferably, in the step of obtaining the block distance, for each of a plurality of neighboring blocks of the current frame classified according to a sequence adjacent to the detected image sample block, Neighboring blocks having the smallest sum of absolute distance or sum of absolute difference (SAD) with respect to the block are regarded as matching blocks, and neighboring blocks in the subsequent stage are blocks detected in the previous step of the detected image sample block or matching block A matching block having a smallest SAD value among the matching blocks in the plurality of steps is selected as a final matching block for the detected image sample block, Lt; / RTI >

Preferably, in the first comparison step, a predetermined maximum block distance is allocated to the detected image sample block when the obtained block distance is equal to or greater than the first threshold value.

Preferably, in the calculating of the intermediate value, the estimated or allocated block distances for the detected image sample blocks are sorted in order of magnitude, and the block distance aligned in the middle of the aligned block distances is detected And estimates the representative block distance for the image sample blocks.

According to another aspect of the present invention, there is provided an apparatus for determining the presence or absence of a user's hand motion or intentional motion during image recognition through shooting, the apparatus comprising: an edge detector for detecting image sample blocks of a previous frame using an edge detection method; A motion vector extractor for comparing the detected image sample blocks with estimated image sample blocks of a current frame to obtain block distances, and comparing the obtained block distances with a first threshold value; And a motion determiner for comparing the obtained block distance with a second threshold when the obtained block distance is smaller than the first threshold in the first comparison step.

The present invention provides the following advantages.

The present invention adopts an edge-based adaptive neighbor block search method to improve the motion presence / absence determination performance in a user selection region-based motion presence / absence determination apparatus using a mobile camera or the like, , There is an advantage that accuracy of motion vector estimation according to neighboring block search method is improved.

In addition, the present invention has an advantage of speeding up the execution of an algorithm through an adaptive search method, reducing the amount of calculation through motion determination using a threshold value, and determining the presence or absence of motion robust against camera shake.

FIG. 1 is a diagram for explaining differences in motion estimation according to the size of a block to be matched,
2 is a block diagram schematically illustrating a configuration of a mobile terminal including a motion determiner for determining whether a user has shaken or intentionally moved according to a preferred embodiment of the present invention;
3 is a diagram showing a detailed configuration of the motion determining unit shown in FIG. 2,
FIG. 4 is a flowchart illustrating a method for determining whether a user has shaken or intentional motion according to a preferred embodiment of the present invention. FIG.
FIG. 5 is a view for explaining the region dividing step and the image sample block detecting step shown in FIG. 4,
FIG. 6 is a diagram for explaining an adaptive 8 neighbor block search method;
FIG. 7 is a diagram for explaining an error of motion estimation according to a sudden motion of a user; FIG.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings. In the following description of the present invention, a detailed description of known functions and configurations incorporated herein will be omitted so as to avoid obscuring the subject matter of the present invention.

Hereinafter, in the embodiments of the present invention, the expression 'motion vector' is used for the sake of understanding. However, the present invention uses only the block distance corresponding to the size of the motion vector to determine whether there is camera shake or intentional motion of the user , It should be noted that the expression of the magnitude of such a motion vector or motion vector may be replaced by the term block distance.

2 is a block diagram of a portable terminal including a motion determiner for determining whether a user has shaken or intentional motion according to a preferred embodiment of the present invention. That is, the present invention divides the motion of the user (or the motion of the subject) displayed on the image through the photographing into the movement due to the hand tremor of the user and the intentional movement of the user, The present invention provides a method for judging resultantly that there is no movement of the user and accurately determining whether or not the user intentionally moves.

The portable terminal 100 includes a camera 110, an image signal processor (ISP) 120, a display unit 130, a wireless communication unit 140, a motion determiner 200, a controller 150, 160). FIG. 1 shows a schematic configuration of the portable terminal 100, but the portable terminal 100 may further include a speaker, a microphone, and a user input device.

The camera 110 forms an image of a subject and detects the formed image as an electric signal. For this, although not shown, the camera 110 includes a lens system having at least one lens to form an image of a subject, a charge-coupled device (CCD) image for detecting an image formed by the lens system as an electric signal Sensors, CMOS (complementary metal-oxide semiconductor) image sensors, and the like.

The image signal processing unit 120 processes a video signal input from the camera 110 according to the control of the controller 150 on a frame unit basis and displays the screen characteristics (size, image quality, resolution, etc.) And outputs the converted image frame.

The display unit 130 displays the image frame input from the image signal processing unit 120 on the screen. The display unit 130 may be a liquid crystal display, a touch screen, or the like. The touch screen displays an image according to the control of the controller 150 and generates a key contact interrupt when a user's input means such as a finger, a stylus pen or the like touches the surface thereof, 150 to the controller 150. The control unit 150 controls the display unit 150 to display the user input information including the input coordinates and the input state. For example, when the character recognition application is being executed by the control unit 150, the display unit 130 may display the image frame input from the image signal processing unit 120 in a preview state on the screen. That is, when executing a mobile-based application such as character recognition, the user shoots a subject in the region of interest while checking the previewed image displayed on the display unit 130 through the camera 110. At this time, it is the input image through the camera 110 that has the greatest influence on the performance of the application, and the application uses the motion information of the subject to shoot when there is no movement of the subject, So that stable performance can be achieved. At this time, the movement of the subject shown in the front and rear input images according to the chronological order includes the movement of the subject in the fixed state of the camera, the shaking of the user in the fixed state of the subject, or the intentional movement.

The radio communication unit 140 receives a radio downlink signal from the public using an antenna and outputs downlink data obtained by demodulating the radio downlink signal to the controller 150. Also, the wireless communication unit 140 modulates the uplink data input from the controller 150 to generate a wireless uplink signal, and wirelessly transmits the generated uplink wireless signal to the public using an antenna. The modulation and demodulation may be performed according to a code division multiple access (CDMA) scheme, a frequency division multiplexing (FDM) scheme, a time division multiplexing (TDM) scheme, Method or the like.

The motion determiner 200 determines whether the user has shaken or intentionally moved from the video frames input from the video signal processor 120 and outputs the determination result to the controller 150. [

The memory 160 stores databases related to images, user information, documents and the like for providing applications of various functions such as a character recognition application and a related graphical user interface (GUI) (Menu screen, idle screen, etc.), operating programs, and the like necessary for driving the display device.

The controller 150 executes an application according to user input information, and the application performs a program operation according to user input information. For example, when a character recognition application is executed and a user determines an area of interest in an image displayed on the display unit 130, the character recognition application performs a character recognition operation. At this time, the motion determination unit 200 determines whether the user has shaken or intentional motion in the image, and outputs the result to the controller 150, so that the character recognition application can recognize the correct character So that the recognition operation can be performed. As described above, the user input information may be input to the controller 150 through a separate user input device such as the display 130 or the keypad.

FIG. 3 is a diagram illustrating a detailed configuration of the motion determining unit 200, and FIG. 4 is a flowchart illustrating a method for determining whether a user has shaken or intentional motion according to a preferred embodiment of the present invention.

The motion determiner 200 includes an edge detector 210, a matching unit 220, a motion vector extractor 230, a median filter 240, and a motion determiner 250, The method includes an image acquisition step S110, a segmentation step S120, an image sample block (or edge block) detection step S130, a first motion vector estimation step S140, a first comparison (or block distance comparison) step (S150), a maximum motion vector allocation step (S160), a second motion vector estimation step (S170), and a second comparison (or motion presence determination) step (S180).

The image acquisition step (S110) is a step in which the edge detector 210 receives an image frame from the image signal processing unit 120.

The region segmentation step S120 may be performed by the edge detector 210 in a region of interest determined in the image displayed on the display unit 130 by the user in the image frame, And dividing the region of interest set according to the default value into a plurality of blocks (or image blocks). This segmentation step S120 is a virtual step, for example, the ROI may be divided into N * N blocks. In this manner, the edge detector 210 processes the image frame in blocks.

5 is a view for explaining the region dividing step and the image sample block detecting step. 5A shows a previous image frame as a whole and an area of interest 310 is set for the previous image frame and the area of interest 310 is divided into 10 * 10 blocks 320 .

The image sample block detection step S130 may include detecting a predetermined number of image sample blocks (or edge blocks) having a high frequency component (i.e., an edge component) among the plurality of blocks, to be. To this end, the edge detector 210 detects image sample blocks including edge components in the plurality of blocks, arranges the detected image sample blocks in descending order of high frequency components, A predetermined number of upper image sample blocks are finally selected. As the edge detector 210, a conventional Sobel edge detector may be used. It is preferable that the number of the image sample blocks to be selected is set to an odd number, that is, 2n + 1 (n = 0, 1, 2, ...) It is for this reason. Through the selection of the image sample blocks, it is possible to prevent the possibility of erroneous block matching due to the absence of the high frequency component as described above. In addition, in order to reduce the amount of calculation, it is preferable that the number of the image sample blocks to be selected is set to five or less.

FIG. 5B shows the current (or later) image frame as a whole, and edge blocks 330 selected for the previous image frame and matching blocks 340 corresponding to the movement of the subject are shown have.

The region segmentation step S120 and the image sample block detection step S130 are performed on the previous image frame, and the subsequent steps are performed on the current image frame. That is, since the present motion detection method is a method of detecting motion of the same object appearing in two forward and backward image frames in chronological order, an image sample block to be subjected to motion estimation is searched in a previous image frame, A block (i.e., a matching block) that is matched (that is, corresponding to the image sample block after the motion) is searched in the current image frame. In other words, the hand shake or intentional motion of the user compares the detected image sample blocks in the previous image frame with the estimated image sample blocks of the current image frame (corresponding to the detected image sample blocks after the motion) .

In the first motion vector estimation step S140, the matching unit 220 performs an adaptive neighboring block search for each of the selected image sample blocks to search for matching blocks corresponding to the image sample blocks , And the motion vector extractor 230 estimates a motion vector for each of the image sample blocks (or each matching block). Unlike the present embodiment, the matching unit 220 may be integrated into the motion vector extractor 230.

In the present embodiment, as the adaptive neighboring block search method, first block matching is performed on eight neighboring blocks that are adjacent to an image sample block initially, and then blocks adjacent to the first matching block are subjected to 2 Adaptive 8 neighboring block search method that gradually broadens the search range in a manner that car block matching is performed. Hereinafter, an adaptive 8 connected neighborhood search method is described for one image sample block, but the same applies to the remaining image sample blocks.

6 is a diagram for explaining an adaptive 8 neighbor block search method.

6A is a diagram illustrating an example of an image block 410 of a previous image frame to be searched and eight primary neighbor blocks 420 of a current image frame neighboring the image sample block 410 (in other words, Lt; / RTI > FIG. 6 (b) shows the retrieved primary matching block 430a among the eight primary neighboring blocks 420. FIG. At this time, the matching block means a block which is recognized as being identical or similar to the image sample block 410. At this time, the matching unit 220 calculates a sum of absolute distance (SAD) value of each of the primary neighboring blocks 420 based on the image sample block 410, The block is regarded as a primary matching block 430a.

Referring to FIG. 6C, the matching unit 220 again selects three secondary neighboring blocks 440a adjacent to the primary matching block 430a, and outputs the image sample block 410 A sum of absolute distance (SAD) value of each of the secondary neighbor blocks 440a is calculated on the basis of the SAD value, and a secondary neighboring block having the smallest SAD value is regarded as a secondary matching block. In this way, the matching unit 220 searches the matching block while gradually widening the search range, and the searching step of the stepwise matching block is performed until the SAD value is smaller than the preset threshold value, Lt; / RTI > Also, the final matching block is determined as a matching block having the smallest SAD value among the matching blocks. Through the adaptive neighbor block search method, the search range of the matching block is widened, and the amount of calculation can be reduced while improving the motion estimation performance.

6 (b) and 6 (c) show the case where the primary matching block 430a is aligned with the image sample block 410 in the row or column direction, and FIGS. 6 (d) and 6 And the matching block 430b is aligned with the image sample block 410 in the diagonal direction. There is no difference in the search method in the case of the two, but in the latter case, the number of the secondary neighboring blocks 440b is larger than the case of the former.

FIG. 6 (b) shows the retrieved primary matching block 430b among the eight primary neighboring blocks 420. FIG. Referring to FIG. 6C, the matching unit 220 again selects five secondary neighboring blocks 440b adjacent to the primary matching block 430b, and selects the edge block 410 as a reference , The SAD value of each secondary neighboring block 440b is calculated, and the secondary neighboring block having the smallest SAD value is regarded as a secondary matching block.

It is necessary to experimentally set the camera shake tolerance (ie, the size of the maximum motion vector allowing camera shake) to be 2 to 8 (in pixel units) for a 640 * 480 resolution image, 5 is preferable.

In the first comparison (or block distance comparison) step S150, the motion vector extractor 230 calculates a distance (i.e., a block distance or a matching distance) between each image sample block and a final corresponding matching block, And comparing the block distance with a maximum distance which is a predetermined first threshold value. In addition, the motion vector extractor 230 estimates (or extracts) motion vectors for the image sample blocks, and compares the estimated motion vector with the maximum distance. At this time, the motion vector corresponds to a displacement vector between the image sample block and the matching block. The maximum distance corresponds to a threshold value of a block distance or a matching distance that should be determined by a user's intentional motion. When the block distance (or the sum of pixel distances as a matching distance value) is greater than or equal to the maximum distance, And determines a predetermined maximum motion vector for the image sample block. The size of the maximum motion vector may be set to coincide with the maximum distance. The primary comparison step S150 is to prevent an error in motion estimation that occurs when intentional motion of the user (or the subject) occurs suddenly. 7 is a diagram for explaining an error of motion estimation according to a sudden motion of a user. FIG. 5A illustrates an image sample block 510 for a previous image frame, FIG. 6B illustrates an image sample block 510 and an estimated matching block 520 for the current image frame, do. As shown, the adaptive neighbor block search method is a local search method rather than a global search method. Therefore, when a subject or a user moves rapidly, a motion estimation error as shown may occur.

Accordingly, when the block distance to a specific image sample block is equal to or larger than the maximum distance that is the first threshold, the predetermined maximum motion vector is assigned (or estimated) to the detected image sample block or the corresponding matching block (or estimated image sample block) Thereby inducing the user to judge that there is an intentional movement. The maximum distance is set to a value larger than the camera shake tolerance. For example, it is desirable to set the maximum distance to 2000 to 4000 when the block size is 8 * 8 (pixel unit), and 2000 to 4000 when the block size is 16 * 16.

In the maximum motion vector allocation step S160, the motion vector extractor 230 allocates a predetermined maximum motion vector to the image sample block or the matching block when the block distance to a specific image sample block is greater than the maximum distance . In this embodiment, it is exemplified that the second motion vector estimation step (S170) is performed after the maximum motion vector allocation step (S160). However, when the block distance to a specific image sample block is larger than the maximum distance, The motion vector estimation step S170 may be omitted and the maximum motion vector may be estimated as a representative motion vector to perform the second comparison (or motion presence / absence determination) step S180. At this time, the number of image sample blocks to which the maximum motion vector is assigned (a threshold corresponding to the total number of the detected image sample blocks or a smaller number of image sample blocks) may be considered.

In the second motion vector estimation step S170, the median filter 240 arranges the motion vectors allocated or estimated for the image sample blocks in order of magnitude, and outputs a motion vector having the median value to the image sample (Or representative) motion vector for the blocks. This step (S160) compensates for the degradation of the precision of the motion estimation as the block size decreases, and improves the reliability of the motion estimation by eliminating the motion vectors that may be erroneously estimated.

In the second comparison (or motion presence / absence determination) step S180, the motion determiner 250 determines whether the motion vector of the current motion vector is greater than a predetermined second threshold If the size of the representative motion vector is less than the predetermined second threshold, it is determined that there is no motion of the user or there is a hand tremor of the user. Otherwise, it is determined that there is an intentional motion of the user.

While the present invention has been particularly shown and described with reference to exemplary embodiments thereof, it will be understood by those skilled in the art that various changes and modifications may be made without departing from the spirit and scope of the invention as defined in the appended claims.

It will be appreciated that the apparatus and method for determining the presence or absence of hand movement or intentional movement of the present invention may be implemented in the form of hardware, software (i.e., program), or a combination thereof. Such a program may be stored in a machine-readable volatile or non-volatile recording medium such as a computer, and the recording medium may be a memory such as a RAM, a memory chip, an integrated circuit or the like, a CD, a DVD, , A magnetic tape, or the like. That is, the method for determining whether the user has shaken or intentional motion of the present invention can be embodied in the form of a program including codes for realizing this. Furthermore, such a program may be electrically transmitted through any medium, such as a communication signal propagated by wire or wireless, and the present invention includes equivalents thereof.

The present invention relates to a mobile terminal having a mobile terminal, a mobile terminal, and a mobile terminal. The mobile terminal includes a camera, a video signal processing unit, a display unit, a wireless communication unit,

Claims (10)

In a character recognition method,
Detecting a sample block in a previous frame of the preview image;
Searching a matching block matching the sample block in a current frame;
Obtaining a first block distance between the sample block and the matching block;
And performing a character recognition operation when the first block distance is equal to or less than a first threshold value. The method according to claim 1,
Executing a character recognition application in accordance with a first user input;
Displaying the preview image;
Further comprising setting a region of interest of the preview image according to a second user input,
Wherein the sample block is included in the region of interest. The method according to claim 1,
Displaying the preview image;
Setting a region of interest of the preview image according to a user input;
Further comprising partitioning the region of interest into a plurality of blocks,
Wherein the sample block is one of the plurality of blocks. The method according to claim 1,
The step of performing the character recognition operation includes:
Performing photographing for character recognition when the first block distance is less than or equal to the first threshold;
And recognizing a character in the photographed image. The method according to claim 1,
Obtaining a second block distance for a next frame of the preview image if the first block distance exceeds the first threshold;
And performing a character recognition operation when the second block distance is less than the first threshold value. The method according to claim 1,
Wherein the first block distance is an intermediate value among a plurality of block distances obtained for the current frame of the preview image. The method according to claim 1,
Comparing a plurality of block distances obtained for the current frame of the preview image with a second threshold greater than the first threshold;
Obtaining a second block distance for a next frame of the preview image if any one of the plurality of block distances is equal to or greater than the second threshold;
And performing a character recognition operation when the second block distance is less than the first threshold value. The method according to claim 1,
When either one of a plurality of block distances obtained for the current frame of the preview image is greater than or equal to a second threshold value larger than the first threshold value or when an intermediate value among the plurality of block distances is smaller than the second threshold value, Determining a second block distance for a next frame of the preview image if the second block distance is greater than a threshold value;
And performing a character recognition operation when the second block distance is less than or equal to the first threshold value,
Wherein the first block distance is an intermediate value among the plurality of block distances. A computer-readable recording medium storing a program for executing the character recognition method according to any one of claims 1 to 8. A portable terminal having a computer-readable recording medium storing a program for executing the character recognition method according to any one of claims 1 to 8.

Images