KR102042131B1 - Method for a video stabilization for real-time optical character recognition (OCR) - Google Patents

Abstract

As a preferred embodiment of the present invention, a method of compensating for shaking a character in an image may include setting a portion of an input frame as an output frame; Determining a target word in the output frame using the predicted position of each word by applying the center coordinates and the OCR technique to the output frame; and the center coordinates of the output frame and the position of the target word. And determining a correction amount based on at least one of the regions of the target word.

Description

Method for stabilizing video when real-time character recognition in terminal {Method for a video stabilization for real-time optical character recognition (OCR)}

The present invention relates to a method for recognizing a character in real time in an image photographed by an image photographing apparatus provided in a terminal. More particularly, the present invention relates to a method of stabilizing an image upon character recognition.

The real-time OCR technique recognizes letters in a region of interest set by a user in an image in real time. In this case, the marker is used to mark a particular word in the center on the preview screen. In a touch type terminal such as a smartphone, the interface is supported to place a marker on a specific word of interest to the user, but when the user's hand shakes, the preview screen displayed in the image may shake.

To solve this problem, various image stabilization techniques have been applied. Image stabilization techniques include optical image stabilization (OIS), digital image stabilization (DIS), and camera stabilization.

In general, a method of recognizing camera shake or rotation using a camera stabilization configuration such as Gimbal and compensating for the camera orientation is used. In addition, a method of predicting camera movement using a gyroscope, a method of stabilizing the entire image by grasping the overall motion of the image, predicting the motion in the three-dimensional space of the camera, and stabilizing the movement of the camera to display the image at a new position Methods of synthesizing have been developed. In addition, in order to predict the movement of the screen or the movement of the camera, objects in the image are tracked and implemented in a manner of compensating for the movement of the tracked object.

KR 10-2007-0093995 A

 In a preferred embodiment of the present invention to provide a screen-centric screen stabilization technique.

In a preferred embodiment of the present invention, a method for compensating for shaking a character in an image may include: setting a part of an input frame input from an image photographing apparatus as an output frame; Determining a target word in the output frame using the predicted position of each word in the output frame by applying the center coordinates and the OCR technique to the output frame; and the center coordinate of the nth output frame. And determining a correction amount based on at least one of the position of the target word and the area of the target word.

로 표시되는 것을 특징으로 한다.In a preferred embodiment of the present invention, the step of determining the target word is to determine the target word of the word having the minimum distance between the center coordinate M (n) of the nth output frame and the center coordinate of each word in the output frame In this case, the position of each word is Rect (k) = (x (k), y (k), w (k), h (k)), and x (k) and y (k) are the starting point of each word. Where w (k) is the horizontal length of each word, h (k) is the vertical length of each word, and the center coordinate of each word is

Characterized in that is represented.

In a preferred embodiment of the present invention, the step of determining the target word targets the word if the center coordinate M (n) of the nth output frame falls within the coordinate range of any word recognized in the nth output frame. In this case, the coordinate range of the word ranges from ((x (k), y (k)) to (((x (k) + w (k)), y (k)), vertical for horizontal. In the case of (x (k), y (k)) to ((x (k), (y (k) + h (k))).

In a preferred embodiment of the present invention, the step of determining the correction amount is characterized in that for determining the correction amount based on the distance between the center coordinates of the n-th output frame and the center coordinates of the target word.

In a preferred embodiment of the present invention, the correction amount is determined to be inversely proportional to the distance between the center coordinate of the nth output frame and the center coordinate of the target word.

According to an exemplary embodiment of the present invention, an n th marker M _n is superimposed and displayed at the center of the n th output frame.

In a preferred embodiment of the present invention, the position of the (n-1) th marker M _{n -1} indicated at the center of the (n-1) th output frame is located within any word in the (n-1) th output frame. In the case of positioning, the word is determined as a target word of the nth output frame.

In one embodiment of the present invention, (n-1) th output displayed in the center of the frame (n-1) th marker M _n, each word in the recognition of the _-1 position is the (n-1) th output frame The target word based on a distance between the position of the (n-1) th marker M _{n -1} and the center of each word recognized in the (n-1) th output frame when located in an empty space between the target words. Characterized in determining.

In a preferred embodiment of the invention, (n-1) th marker M _{n -1} of the position and the (n-1) th output and the center distance of the smallest word of each word recognized in the frame n-th It is characterized by determining the target word of the output frame.

According to another preferred embodiment of the present invention, a step of setting a portion of the input frame input from the image photographing apparatus as an output frame; Determining a target word in the output frame using the center coordinates of the output frame and the position of each word recognized in the output frame; and a distance between the center coordinate of the nth output frame and the center coordinate of the target word. Determining an inversely corrected amount; Moving the set output frame by the correction amount; And cropping and displaying the moved output frame in the input frame.

As an exemplary embodiment of the present invention, the method may further include displaying the marker M 'overlapping the center point of the moved output frame.

According to another preferred embodiment of the invention, the step of setting a portion of the input frame input from the image photographing apparatus equipped with a CMOS sensor as an output frame; When rolling shutter distortion occurs in the input frame, performing the rolling shutter distortion correction; Determining a target word in the output frame using the center coordinates of the output frame and the position of each word recognized in the output frame; and a distance between the center coordinate of the nth output frame and the center coordinate of the target word. Determining an inversely corrected amount; Moving the set output frame by the correction amount; And cropping and displaying the moved output frame in the input frame.

According to another preferred embodiment of the present invention, an image recording apparatus for performing shake correction when recognizing characters in an image includes an output frame setting unit for setting a part of an input frame input from the image photographing apparatus as an output frame; A target word determination unit determining a target word in the output frame by using the position of each word in the output frame predicted by applying the center coordinates and the OCR technique of the output frame to the output frame; And a correction unit determining a correction amount based on at least one of a center coordinate, a position of the target word, and an area of the target word.

According to another preferred embodiment of the present invention, an image recording apparatus for performing shake correction when recognizing characters in an image includes an output frame setting unit for setting a part of an input frame input from the image photographing apparatus as an output frame; A target word determination unit that determines a target word in the output frame using the center coordinates of the output frame and the position of each word recognized in the output frame; and a center coordinate of an nth output frame and a center coordinate of the target word. A correction unit which determines an amount of correction inversely proportional to a distance between and moves the set output frame by the amount of correction; And a display unit for cropping and displaying the moved output frame from the input frame.

According to another preferred embodiment of the present invention, the image recording apparatus for performing image stabilization during character recognition in the image output frame setting unit for setting a portion of the input frame input from the image photographing apparatus equipped with a CMOS sensor ; A distortion correction unit configured to perform the rolling shutter distortion correction when the rolling shutter distortion occurs in the input frame; A target word determiner configured to determine a target word in the output frame using the center coordinates of the output frame and the position of each word recognized in the output frame; and a center coordinate of an nth output frame and a center coordinate of the target word. A correction unit for determining a correction amount inversely proportional to a distance between the two and moving the set output frame by the correction amount; And a display unit for cropping and displaying the moved output frame from the input frame.

In a preferred embodiment of the present invention, a technique for performing stabilization around a character rather than an image in a shaking image is proposed. When there are a plurality of letters or words on the screen, the user's intention is automatically detected and the screen is stabilized around the letters or words desired by the user. As a result, the problem that the user's responsiveness is overlooked in the screen stabilization or image stabilization technique is solved, so that the user's small intentional movement or the user's responsiveness can be quickly reflected.

As a preferred embodiment of the present invention, the method of performing image stabilization during image recognition in the image by adjusting only the crop (crop) range of the output frame to output based on the current input frame does not require excessive operation and shakes quickly There is an effect that can correct.

As a preferred embodiment of the present invention, when a rolling shutter distortion occurs in an input frame by using an image photographing apparatus equipped with a CMOS sensor, a shake correction is performed during letter recognition of the present invention after supplementing a rolling shutter distortion generally applied. In this case, even when using an image photographing apparatus equipped with a CMOS sensor, there is an effect of improving character recognition.

1 (a) to (c) shows an example of a magnifying glass app implemented in a smartphone.
2 illustrates a problem of a picture stabilization technique using a low band filter.
FIG. 3 is a diagram illustrating an internal configuration of an image photographing apparatus that performs shake correction when recognizing letters in an image according to an exemplary embodiment of the present invention.
4 illustrates a relationship between an input frame and an output frame as a preferred embodiment of the present invention.
FIG. 5 is a diagram for one example of correcting a shake when a character is recognized in an image as an exemplary embodiment of the present invention.
6 illustrates an example of setting a target word in an output frame as a preferred embodiment of the present invention.
FIG. 7 illustrates an example of comparing the example corrected according to the present invention with the example corrected according to a conventional method when a shaking occurs in character recognition in an input image.
8 to 9 illustrate a flowchart for correcting when a shake occurs in character recognition in an input image as an exemplary embodiment of the present invention.

1 (a) to (c) shows an example of a magnifying glass app implemented in a smartphone.

Figure 1 (a) is a "bicycle" (110) of the words "car" 120, "bicycle" 110, "train" 130 displayed on the smartphone screen by the user using a smartphone magnifying glass app (110) ) Shows an example in which the enlarged word "bike" 110a enlarged by enlarging the word) is superimposed and displayed on a smartphone screen.

In the input image, the distance between the word "train" 130 and the edge of the screen is d1 (111) in FIG. 1 (a), d2 (112) in FIG. 1 (b), and FIG. 1 (c). In the case of) it can be seen that d3 (113).

The user enlarges the word "bike" 110 as shown in FIG. 1 (a), and enlarges the word "car" 120 which is the left word of the "bike" 110 as shown in FIG. 1 (b). By looking at the enlarged word "car" (120a), and then moving the screen of the smartphone again to enlarge the "bike" 110 as shown in the example of Figure 1 (c) to see the enlarged word "bike" (110c). .

Conventionally, when the user moves the screen of the smartphone as shown in Figs. 1 (a) to (c), it is not determined whether the intention of the user's movement is intentional or unintentional movement. When the screen of the smartphone moves simply to stabilize the screen, the movement of the smartphone is converted into a smooth path by passing through a low pass filter (LPF) based on the degree of movement of the smartphone or location information of the smartphone. It was common to show a smartphone image to a user.

In this case, the motion of the user moving from Fig. 1 (a) to Fig. 1 (b) and then back to Fig. 1 (c) by applying a technique of stabilizing the shaking by judging the small movement of the user as the shaking of the unwanted screen. There was a problem that the smartphone screen is not reflected as it is.

Referring to Figure 2, looks at the problem that occurs when the screen is stabilized by using a low-band filter. In general, since the shaking of the screen has a high frequency component, a low band filter is applied to reduce the shaking. When the input signal 200 is passed through the low pass filter, it is possible to obtain stable image signal values 201 and 202 in which small residual components are efficiently removed.

However, in the case of the low band filter, low frequency filtering is performed using only information related to past or present time points acquired from an image, and thus, when a sudden movement occurs in an image photographing apparatus such as a camera, the stable image signals 201 and 202 are shown in FIG. 2. Is obtained, but there is a problem that can not follow the sudden change R1 (210), R2 (220).

Even when the weak low frequency filtering is performed, a difference in response time R2 220 occurs between the input signal 200 and the low frequency filtering image signal 201, and when the low frequency filtering is performed in general, the input signal 200 and the low frequency filtering are performed. A very large difference in response time R2 220 occurs between the image signals 202 subjected to low frequency filtering.

In other words, by using a low-pass filter to remove a small blur component can ensure a stable path, but there is a problem that can not follow the change, such as sudden camera movement or the user does not detect a small intentional movement. This phenomenon occurs when the user moves the smartphone, the screen is not updated immediately.

In a preferred embodiment of the present invention as described above, when using a low-band filter or when applying a conventional image stabilization or image stabilization technique, the response time is too late, or the high frequency component of the input signal is removed excessively I would like to.

FIG. 3 is a diagram illustrating an internal configuration of an image photographing apparatus for performing shake correction when recognizing characters in an image according to an exemplary embodiment of the present invention.

The image capturing apparatus 300 may include an output frame setting unit 310, a target word determiner 320, and a correction unit 330, and may further include a display unit 340.

In the present invention, the image capturing apparatus may be interpreted as including a terminal, a mobile phone, a smart phone, a smart watch, a tablet, a notebook, a computer, a handheld device, a wearable device, and the like.

In addition, the image photographing apparatus includes an optical unit for inputting an optical signal from a subject, an imaging device for converting an optical signal input through the optical unit into an electrical signal, a noise reduction process for the electrical signal provided from the imaging device, a conversion process for a digital signal, and the like. An input signal processor for performing signal processing of the motor, a motor for driving the optical unit, and a driver for controlling the operation of the motor. In addition, a user input unit for inputting a user's operation signal, SDRAM for temporarily storing data of input image, data for arithmetic processing, processing result, flash memory, SD / CF / SM card, etc. as a recording device for storing image file. It may be further provided. In addition, as an imaging device included in the image photographing apparatus, a CMOS (Complementary Metal Oxide Semiconductor) sensor array, a charge coupled device (CCD) sensor array, or the like may be used.

Each configuration of the image capturing apparatus 300 illustrated in FIG. 3 has the following characteristics.

The output frame setting unit 310 is implemented to set a part of the input frame input from the image photographing apparatus as an output frame. Referring to FIG. 4, the output frame setting unit 310 may set the (n−1) th output frame 410 or the n th output frame 420 in the input frame 400, respectively.

The target word determiner 320 may determine the target word in the output frame by applying the OCR technique to the output frame using the position of each word in the predicted output frame, the region of each word, and the center coordinate of the output frame. The target word determiner 320 may determine the target word using only the current output frame information or may further determine the target word using the previous output frame information.

As an exemplary embodiment of the present invention, the target word determiner 320 determines the target word using only the current output frame information as follows.

로 표시될 수 있다. The target word determiner 320 determines a word having a minimum distance between the center coordinate M (n) of the nth output frame and the center coordinate of each word in the output frame. The position of each word of the nth output frame is Rect (k) = (x (k), y (k), w (k), h (k)), and x (k) and y (k) are each word. Where w (k) is the width of each word, h (k) is the length of each word, and the center coordinates of each word are

It may be represented as.

Referring to the exemplary embodiment of FIG. 6, a part of the input image 600 may be set as the output image 610. Each word, “new”, “car”, “bicycle” and “train” in the output image 610 may be recognized by the OCR technique, and the position of each recognized word may be predicted.

In one preferred embodiment of the present invention, the input image 600 uses an absolute coordinate system. If the reference image 600, the "bicycle" (601) is close to the center of the input image 600, if the reference to the output image 610, "new" 611 is the output image ( 610 may be calculated close to the center. In an exemplary embodiment of the present invention, the target word determiner 320 may determine “new” 611, not “bicycle” 601, based on the distance from the center of the output image 610 as the target word.

In this case, the correction unit 330 may perform correction by moving the output screen by a value inversely proportional to the distance between the center 612 of the target word “new” 611 and the center of the output image 610 in the output image. have. In another embodiment of the present invention, the correction amount may be determined to be inversely proportional to the square of the distance between the center of the target word and the center of the output image.

When the correction amount is determined, the display unit 340 moves the screen image by the correction amount, and then crops the output image in which the correction amount is reflected in the input image and displays it on the display. In this case, the n th marker M _n may be superimposed on the center of the n th output frame and displayed together on the display.

The greatest difference between the present invention and conventional stabilization techniques occurs here. In the conventional image stabilization technique, since the motion of the image or the movement of the camera must be predicted, it is necessary to analyze the continuous images or the images photographed at different times. For example, analysis between the nth image and the n-1th image is necessary. However, in one embodiment of the present invention, there is a difference that shake correction can be performed using only the n-th output frame 610.

As another example, the target word determiner 320 may determine the target word when the center coordinate M (n) of the nth output frame falls within the coordinate range of any word recognized by the nth output frame. In this case, the coordinate range of the word ranges from ((x (k), y (k)) to (((x (k) + w (k)), y (k)) for the horizontal and vertical ((x (k), y (k)) to ((x (k), (y (k) + h (k))). In this case, the correction unit 330 is configured to display the output image. Correction may be performed by moving the output screen by a value inversely proportional to the distance between the center 612 of the target word “bird” 611 and the center of the output image 610. In addition, the correction unit 330 may perform correction. The amount of correction may be determined in consideration of the size of the word "611".

Referring to FIG. 6, when the area of the word “new” is (x1, y1) to (x1 + w1, y1 + h1), the center (not shown) of the output image 510 is the area of the word “new”. When included within (x1, y1) to (x1 + w1, y1 + h1), the word “new” can be recognized as a target word. This case takes the size of the word "new" into account. For example, based on at least one of the area amount calculated based on the horizontal w1 * length h1 of the word “new” 611 or the distance between the center 612 of the “bird” 611 and the center of the output image 610. The amount of correction can be determined with.

That is, according to the distance between the earth and the object, such as the universal gravity, the position can be corrected by a model in which an arbitrary word recognized in the output image pulls the center point of the output image. The w1 x h1 corresponding to the size of the word "new" 611 corresponds to the mass, and the distance between the center of the word "new" 612 and the center of the output image 610 corresponds to the distance between the earth and the object. This allows us to define forces like the universal gravity. Depending on the magnitude of the calculated force, the position change over time can be calculated. In this model, the closer the word is to the center, the stronger the force, and the farther away the effect is. Therefore, when the distance between the word and the target word is close to the center, the target word is always positioned at the center of the output screen, and the user may feel that the screen is stabilized. When the user moves the target word more than a certain size from the center of the output image due to a sudden movement, the force of interaction between the word and the output image is reduced, leaving the gravity of the word.

As another preferred embodiment of the present invention, the target word determiner 320 may determine the target word using only previous output frame information and current output frame information.

Referring to FIG. 4, the position 411 of the (n-1) th marker M _{n -1} displayed at the center of the (n-1) th output frame 410 is the (n-1) th output frame 410. If it is located inside any word in the word, the word may be determined as the target word of the nth output frame.

In another preferred embodiment, (n-1) th output frame 410 is displayed in the center of (n-1) _n-th marker M position 411 of _-1 (n-1) th output of the frame (410 In the space between each word recognized within the), the center of each word recognized within the position 411 of the (n-1) th marker M _{n -1} and the (n-1) th output frame. The target word can be determined based on the distance of.

In another preferred embodiment, the distance between the position 411 of the (n-1) th marker M _{n -1} and the center of each word recognized in the (n-1) th output frame 410 is the minimum. The word may be determined as the target word of the nth output frame.

FIG. 5 is a diagram for one example of correcting a shake in an output image through correction by a correction unit.

When the range of the output image in the nth input image 500a is set to be a distance d1 (501) between the word "train" and one side of the screen (501), and there is no shaking, the nth output image (510) ), The distance d3 503 between the word "train" and one side of the screen is substantially the same.

However, when the image is minutely moved to the right due to the minute movement of the image photographing device when the n + 1th input image 500b is photographed, the distance between the word "train" and one side of the screen is d1 (502). Becomes smaller than 501. In a preferred embodiment of the present invention, the output image is set in the n + 1th input image 500b, the target word is set based on the center of the set output image and the position of each word in the output image, and the target word. The correction is performed based on the center of the center and the center of the set output image to output the n + 1 th output image 520. In this case, the amount of correction can also be determined to be equal to d3 (503).

In an exemplary embodiment of the present invention, in FIG. 5, the n + 1 th output image 520 determines the correction amount based on only the n + 1 th input image 500b and is displayed on the n + 1 th output image 520. Train ”and the distance d4 504 of one side of the screen may determine the correction amount to be larger than d2 502.

As another preferred embodiment of the present invention, in FIG. 5, the n + 1 th output image 520 may determine the correction amount in consideration of both the n th output image 510 and the n + 1 th input image 500b. .

In this case, when the center point of the n + 1 th output image 520 is M _{n +1} and the center point of the n th output image 510 is M _n , the center point and the n th output of the n + 1 th output image 520 are output. The center point of the image 510 has the following relationship.

M _{n +1} = M _n + update (n), update (n) is the correction amount of the nth output image 510

update (n) may be determined in consideration of the position of the target word, the size (region) of the target word, and the center Mn of the nth output image 510.

FIG. 7 illustrates an example of comparing the example corrected according to the present invention with the example corrected according to a conventional method when a shaking occurs in character recognition in an input image.

A second input frame 701, a 16th input frame 702, a 26th input frame 703, and a 38th input frame 704 of the input image 700 are illustrated. The markers 701a, 702a, 703, and 704a, which represent the center points of the respective input frames 701 to 704, are constantly at the center of the word "have", but show a slight movement between the input frames.

If there is a slight movement in the image photographing apparatus, in the output image 720, which is corrected according to the existing method, the second output frame 721, the 16th output frame 722, the 26th output frame 723 and The center points 721a, 722a, 723a, and 724a of the 38th output frame 724 all leave the word "have" and are located in the empty space between the words. It can be seen that it is inefficient in compensating for the minute movement occurring in the image including the characters photographed by the image photographing apparatus.

The output image 710 correcting the input image 700 by applying the shake correction method in the image recognition according to the present invention includes a second output frame 7141, a 16th output frame 712, and a 26th output frame 713. ) And the center point of the 38th output frame 714 are located at the center points 711a, 712a, 713a, and 714a of the word "have", so that the difference between the centers of each of the frames 711 to 714 is substantially almost different. It is difficult. That is, even when a shake occurs when a character is recognized in the input image, the image pickup device can stably compensate for the shake.

8 to 9 illustrate a flowchart for correcting when a shake occurs in character recognition in an input image as an exemplary embodiment of the present invention.

8 is a flowchart illustrating a case in which shaking occurs when character recognition in an input image is performed according to an exemplary embodiment of the present invention.

The output frame setting unit sets a portion of an input frame input from the photographing apparatus as an output frame (S810). The target word determination unit determines the target word using the center coordinates of the output frame set by the output frame setting unit and the position of each word recognized in the output frame (S820). In a preferred embodiment of the present invention, by applying the OCR technique in real time to predict or recognize the position of each word in the output frame, in addition to the words, letters, characters in the image using a variety of word detection techniques Etc. can be detected.

The correction unit determines a correction amount inversely proportional to the distance between the center coordinate of the nth output frame and the center coordinate of the target word (S830), and moves the set output frame by the amount of correction (S840). The display unit crops and displays the moved output frame from the input frame (S850).

FIG. 9 is a flowchart illustrating a case in which shaking occurs when a character is recognized in an input image in an image photographing apparatus using CMOS according to an exemplary embodiment of the present invention.

In the output frame setting unit, a part of the input frame input from the image photographing apparatus equipped with the CMOS sensor is set as an output frame (S910). In a preferred embodiment of the present invention, when the rolling shutter distortion occurs in the input frame due to the CMOS sensor, after performing the rolling shutter distortion correction (S920), the shake correction generated during character recognition in the input image is performed (S930 to S960). ). For the shake correction generated when the character in the input image is recognized, refer to steps S820 to S850 of FIG. 8.

The invention can also be embodied as computer readable code on a computer readable recording medium. The computer-readable recording medium includes all kinds of recording devices in which data that can be read by a computer system is stored. Examples of computer-readable recording media include ROM, RAM, CD-ROM, magnetic tape, floppy disks, optical data storage devices, and the like. The computer readable recording medium can also be distributed over network coupled computer systems so that the computer readable code is stored and executed in a distributed fashion.

The embodiments of the present invention described above and illustrated in the drawings should not be construed as limiting the technical spirit of the present invention. The protection scope of the present invention is limited only by the matters described in the claims, and those skilled in the art can change and change the technical idea of the present invention in various forms. Therefore, such improvements and modifications will fall within the protection scope of the present invention if it is obvious to those skilled in the art.

Claims (16)

Setting a part of an input frame input from an image photographing apparatus as an output frame;
Determining a target word in the output frame using the predicted position of each word in the output frame by applying the center coordinates and the OCR technique to the output frame; and
Determining a correction amount based on at least one of a center coordinate of an output frame, a position of the target word, and an area of the target word;
And the correction amount is determined to decrease when the distance between the center coordinate of the output frame and the center coordinate of the target word increases. The method of claim 1, wherein the determining of the target word
The target word is a word having a minimum distance between the center coordinate M (n) of the nth output frame and the center coordinate of each word in the output frame. In this case, the position of each word is Rect (k) = (x (k ), y (k), w (k), h (k)), x (k), y (k) is the starting point of each word, w (k) is the width of each word, h (k) Represents the vertical length of each word, and the center coordinate of each word is

Shake correction method when recognizing the character in the image, characterized in that the display. The method of claim 1, wherein the determining of the target word
If the center coordinate M (n) of the nth output frame falls within the coordinate range of any word recognized by the nth output frame, the target word is determined for the word, in which case the coordinate range of the word is horizontal ( (x (k), y (k)) to (((x (k) + w (k)), y (k)), for vertical ((x (k), y (k)) to (( Image stabilization method for character recognition in the image, characterized in that x (k), (y (k) + h (k))). The method of claim 1, wherein the determining of the correction amount
Shake correction method for character recognition in the image, characterized in that for determining the correction amount based on the distance between the center coordinates of the output frame and the center coordinates of the target word. delete The method of claim 1,
An image stabilization method for character recognition in an image, wherein an n th marker M _n is superimposed and displayed at the center of the n th output frame. The method of claim 6,
(n-1) if the output of the second displayed in the center of the frame (n-1) th marker M _{n -1} position located within the random word in the (n-1) th output frame, the second word n Image stabilization method for character recognition in the image, characterized in that for determining the target word of the output frame. The method of claim 6,
When the position of the (n-1) th marker M _{n -1} indicated in the center of the (n-1) th output frame is located in the empty space between each word recognized in the (n-1) th output frame, The target word is determined based on a distance between the position of the (n-1) th marker M _{n -1} and the center of each word recognized in the (n-1) th output frame. Image stabilization method for character recognition. The method of claim 8,
Determining a word having a minimum distance between the position of the (n-1) th marker M _{n -1} and the center of each word recognized in the (n-1) th output frame as the target word of the nth output frame Image stabilization method of character recognition in the image. Setting a part of an input frame input from an image photographing apparatus as an output frame;
Determining a target word in the output frame using the center coordinates of the output frame and the position of each word recognized in the output frame; and
Determining an amount of correction inversely proportional to a distance between a center coordinate of an output frame and a center coordinate of the target word;
Moving the set output frame by the correction amount;
And displaying the shifted output frame by cropping the input frame from the input frame. The method of claim 10,
And overlaying the marker M 'on the center point of the moved output frame to display the shake correction method. Setting a part of an input frame input from an image photographing apparatus equipped with a CMOS sensor as an output frame;
When rolling shutter distortion occurs in the input frame, performing the rolling shutter distortion correction;
Determining a target word in the output frame using the center coordinates of the output frame and the position of each word recognized in the output frame;
Determining an amount of correction inversely proportional to a distance between a center coordinate of an output frame and a center coordinate of the target word;
Moving the set output frame by the correction amount; and
And displaying the shifted output frame by cropping the input frame from the input frame. An output frame setting unit that sets a portion of an input frame input from the image photographing apparatus as an output frame;
A target word determination unit for determining a target word in the output frame using the position of each word in the output frame predicted by applying the center coordinates and the OCR technique of the output frame to the output frame; And
And a correction unit determining a correction amount based on at least one of a center coordinate of an output frame, a position of the target word, and an area of the target word.
And the correction amount is determined to decrease when the distance between the center coordinate of the output frame and the center coordinate of the target word increases. An output frame setting unit that sets a portion of an input frame input from the image photographing apparatus as an output frame;
A target word determination unit which determines a target word in the output frame by using the center coordinates of the output frame and the position of each word recognized in the output frame; and
A correction unit which determines a correction amount in inverse proportion to a distance between a center coordinate of an output frame and a center coordinate of the target word and moves the set output frame by the correction amount;
And a display unit for cropping and displaying the shifted output frame in the input frame. An output frame setting unit for setting a portion of an input frame input from an image photographing apparatus having a CMOS sensor as an output frame;
A distortion correction unit configured to perform the rolling shutter distortion correction when the rolling shutter distortion occurs in the input frame;
A target word determination unit which determines a target word in the output frame by using the center coordinates of the output frame and the position of each word recognized in the output frame; and
A correction unit which determines a correction amount in inverse proportion to a distance between a center coordinate of an output frame and a center coordinate of the target word and moves the set output frame by the correction amount;
And a display unit for displaying the moved output frame by cropping the input frame from the input frame. A computer-readable recording medium having recorded thereon a program for performing the method according to any one of claims 1 to 4.

Images