KR100871012B1 - Method for ground shadow reduction in sport video - Google Patents

Abstract

The present invention relates to a method for reducing ground shadows in athletics video. The present invention relates to a method for reducing ground shadows for improving the viewing experience of athletics video. The present invention relates to ground shadows using RGB values in each image frame of athletics video. Determining whether or not, dividing an image frame having the ground into n × n blocks, constructing a ground block map, detecting shot boundaries by checking temporal discontinuity of the constructed ground block map; Determining a distant or non-distant shot at the detected shot boundary; determining if there is a shadow region in the corresponding image frame if determined to be the far shot; and if the shadow region exists, YU in the corresponding image frame. Extracting a shadow area using the value, and extracting Classifying the first and second shadow areas according to the intensity of the shadow components within the shadow area, and adjusting the brightness according to the intensity of the classified first and second shadow areas to reduce the shadow components. It has the effect of providing a better viewing experience.

Athletics, Watching Experience, Shadow, Ground Block Map

Description

How to reduce ground shadows in athletics video {METHOD FOR GROUND SHADOW REDUCTION IN SPORT VIDEO}

1 is a diagram illustrating three types of shot types present in athletic video, including (a) long shots, (b) medium shots, and (c) close shots.

FIG. 2 is a view showing an image in which shadows are cast in the ground existing in the athletic video. FIG.

3 is a general flowchart illustrating a method for reducing ground shadows in athletics video according to an embodiment of the present invention.

4 is a view showing a state after filling a small hole present in the ground block map in the present invention, (a) is a ground block map, (b) is a state after filling the hole.

5 is a view showing a state of each channel of the ground image having a shadow area in the present invention, (a) the original image, (b) Y channel, (c) U channel, (d) V channel.

6 is a diagram illustrating a histogram for each channel of a region detected as ground in an image in which shadow exists in the present invention, (a) Y channel, (b) U channel.

FIG. 7 is a diagram illustrating an image in which shadow exists in the present invention and a shadow area on the extracted ground. FIG.

8 is a view for comparing the results of performing the ground shadow reduction method in the athletic video according to an embodiment of the present invention, (a) Y histogram of the original image, (b) Y histogram of the result image.

9 is a view illustrating a process of processing a sports video according to the present invention, wherein (a) an original image in which a shadow exists, (b) an image extracted from a shadow region, and (c) an image of reduced shadow.

The present invention relates to a method for reducing ground shadows in athletic video. More specifically, the present invention relates to a shadow region extracted by finding a shadow region in real time through a quick and simple method in a fast motion image such as in field sports. The present invention relates to a method of reducing ground shadows in athletics videos that can provide a better viewing experience to a user by performing shadow reduction for the user.

Recently, with the development of digital and communication technologies, the multimedia environment has reached a level capable of providing a desired service to users without restriction of time and place.

Therefore, up to now, guaranteeing the Quality of Service (QoS) has been an important issue as a measure for providing a stable service to users, and for this purpose, many efforts have been made to provide a higher quality service without interruption in a given bandwidth. Have been.

However, due to the recent rapid development of digital image technology and high expectations of consumers, it provides viewers with the convenience and pleasant experience of watching TV beyond simply guaranteeing quality of service (QoS), which focuses on clear picture quality and seamless transmission. There is a need for display technology that can give.

Now, as the marketing industry pursues "customer satisfaction" beyond the "customer satisfaction", the multimedia content service industry also strives to provide a more pleasant and satisfying environment for users beyond the concept of quality of service (QoS). This has come to the point. Thanks to this recognition, the term quality of experience (QoE) has recently emerged, in which the concept of quality of service (QoS) has been expanded.

Along with such a paradigm change, researches to improve viewing experience or viewing convenience of viewers in the digital image field have been actively conducted.

1 is a diagram illustrating three types of shot types present in a sports video, where (a) is a long shot, (b) is a medium shot, and (c) is a short shot.

Referring to FIG. 1, in the case of a soccer game, which is a representative field sport, sizes of objects present in a field are different depending on the type of shot to be shot. As shown in (c) of FIG. 1, there is no big problem in the case of close-up photography, but as in (a) of FIG. 1, the objects (eg, the player, the ball, etc.) in the screen are very small in the case of the long-distance photography. Often the content is difficult to understand.

In order to solve this problem, a study was attempted to enlarge only a region of interest (ROI) of which users are mainly interested. That is, the conventionally proposed method automatically divides a soccer video into shots that need to extract ROI and shots that do not need to automatically extract ROI for the frames of shots that need to extract ROI. It is an intelligent display method that enlarges and displays.

In this way, the intelligent display method according to the related art does not show the score box in a small screen when watching a soccer game on a mobile terminal, or because the score box does not always exist at a point in time desired by the user. Attempts have been made to find only the part corresponding to the score box and to enlarge it, or to show it at a point in time desired by the user by storing previously stored information even though it does not exist on the actual screen.

As described above, research on improvement of problems that hinder the comfortable viewing environment that may occur even in general TV viewing as well as an uncomfortable viewing experience caused by the small size of the display device needs to be progressed to improve the viewing experience.

For example, FIG. 2 is a view showing a shadow cast in a ground existing in an athletic video. In the case of field sports in a daylight with intense sunlight, a shadow cast on the ground from a stands is viewed by users. You will feel a lot of inconvenience.

That is, in the shadow area, the brightness of the object is generally dark and similar, so it is not easy for users to distinguish it. Therefore, the reduction of the shadow component not only provides a more comfortable viewing environment for the user, but also additionally facilitates object tracking or extraction of the region of interest in the shadow region.

Recently, various studies have been attempted to improve shadow areas, but generally, studies on extraction and removal of shadow areas for still images, shadow areas for video only, or removal through shadow area extraction are performed. It is just a study of low-rate images.

However, field sports competition is required in a very fast and robust shadow improvement method because the movement of the object while moving in real time and fast.

The present invention has been made to solve the above-mentioned problems, an object of the present invention is to find a shadow region extracted in real time through a quick and simple method in the image of the fast movement, such as in field sports, for the extracted shadow region By performing shadow reduction, a method of reducing ground shadow in a sports video that can provide a better viewing environment to a user is provided.

In order to achieve the above object, a first aspect of the present invention provides a method of reducing ground shadows for improving a viewing experience of a sports video, the method comprising: (a) using RGB values in each image frame of the sports video; Determining whether a ground exists; (b) constructing a ground block map by dividing an image frame having the ground into n × n blocks; (c) detecting shot boundaries by checking temporal discontinuity of the constructed ground block map; (d) determining a distant or non-distant shot at the detected shot boundary; (e) if it is determined in the step (d) that the far shot is made, determining whether a shadow area exists in the corresponding image frame; (f) extracting the shadow area using the YU value in the corresponding image frame when the shadow area exists in the step (e); And (g) classifying the first and second shadow areas into the first and second shadow areas according to the intensity of the shadow components in the extracted shadow area, and adjusting the brightness according to the intensity of the classified first and second shadow areas to reduce the shadow components. To provide a method for reducing ground shadows in athletics video comprising the step of.

According to a second aspect of the present invention, there is provided a recording medium on which a program for executing the method of reducing the ground shadow in the above-described athletic game video is recorded.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings. However, embodiments of the present invention illustrated below may be modified in many different forms, and the scope of the present invention is not limited to the embodiments described below. The embodiments of the present invention are provided to more fully explain the present invention to those skilled in the art. Meanwhile, in the following embodiment, a soccer game video is described as an object.

First, in the case of non-distance shots such as close-ups or mid-range shots in a soccer game video, the objects in the game are already magnified and displayed on the screen, so that viewers often feel uncomfortable due to shadows.

However, in the case of a long shot as described above, when a shadow exists, the viewer may feel uncomfortable in properly discriminating an object or determining a situation. Therefore, it is necessary to classify the far shot from the shots received before detecting the shadow area and performing the shadow reduction.

After performing the shot classification, the shadow detection may be performed on the shot having the shadow among the shots determined as the far shot, and the improved shadow image may be provided by reducing the detected shadow component.

3 is an overall flowchart illustrating a method for reducing ground shadows in a sports video according to an embodiment of the present invention.

Referring to FIG. 3, a method of reducing ground shadows in a sports video according to an embodiment of the present invention, first, ground using RGB (Red, Green, Blue) values in each image frame of a specific sports video. In operation S100, a ground block map is constructed by dividing the video frame having the ground into n × n blocks.

Next, after detecting the shot boundary by checking the temporal discontinuity of the ground block map constructed in step S200 (S300), a far or non-distant shot is determined at the detected shot boundary (S400).

Subsequently, when it is determined in the step S400 that the long shot is determined, it is determined whether a shadow region exists in the corresponding image frame (S500). If the shadow region exists, YU (Y: luminance signal, U: luminance signal in the corresponding image frame) exists. The shadow region is extracted using the difference between the and red components) (S600).

Next, the first and second shadow areas are classified according to the intensity of the shadow component in the extracted shadow area (S700), and the brightness is adjusted by adjusting the brightness according to the intensity of the classified first and second shadow areas. Reduce (S800).

When the ground shadow reduction method in the athletics video according to an embodiment of the present invention configured as described above is described in detail step by step.

S100: Determination of the presence of ground

In football video analysis, it is very important to determine the ground pixels under various environments. This is very important not only for detecting edges of shots and determining classification of shots, but also for determining shadow areas on the ground. The color of the ground has a slightly different color depending on the football video, but it can be seen that g> r> b generally has a relationship.

Therefore, the ground discovery in the present invention, that is, the expression for determining whether the existence can be expressed as Equation 1 below.

Here, r, g, and b are RGB values at (x, y) coordinates in the video frame and have a value in the range of 0 to 255. In the (x, y) coordinates, G (x, y), which indicates whether it is ground, has a value of 1 when it belongs to ground, or 0 otherwise.

When I (x, y) is the intensity value of the (x, y) coordinate, the ground is classified into three types according to the I (x, y) value. First, when the intensity is above 200 (CASEⅠ), very bright ground; second, when the intensity is above 50 and below 200 (CASE II); and when the intensity is below 50, third, when the intensity is below 50 ( CASEIII) Classify each into dark ground.

In addition, g> r> b relationship, which is the basic characteristic of ground color, is used, and for each exceptional situation that may be mistaken for grass, weights are assigned to each r, g, and b value as shown in Equation 1 above. It was.

For example, in the case of the yellow uniform has a r, g, b value relationship similar to the ground color, but the difference between g and b value is very large, the weight value (Equation 1) is determined through observation, In the case of turf that has a red color due to the change of turf color, it has r> g> b relationship even though it is ground, so to compensate for this, set 95 as the g value and 85 as the r value. The differences were compared by multiplying the weight values by the r and g values. Then, the condition is adjusted according to each brightness so that the ground can be distinguished regardless of the brightness.

S200: Ground Block Map Construction Steps

In the present invention, for fast shot boundary detection and shot classification, the ground block map GB (i, j) is constructed by dividing the entire frame into 16 × 16 blocks. To define GB (i, j), first, B _ij , a set of pixels in one block, is defined as in Equation 2 below.

The ground block map GB (i, j) is defined as in Equation 3 below using Equation 2 above.

In the present invention, as a result of testing ground block samples obtained through a total of 20 different soccer videos, it can be seen that the number of pixels determined as ground in the ground block is about 50% or more, indicating that the ground block is a reference value (T). _Ground ) to 0.5. Of course, 50%, which is a reference value for determining the ground block, can be changed as much as the set value. The resulting ground block map is as shown in Fig. 4A.

On the other hand, if the ground block map is formed, an object block exists on the ground by a player, a ball, or the like as shown in FIG. In the present invention, the object block existing on the ground is defined as a hole in order to distinguish the ground area from the outside of the ground, such as a spectator seat, and to use it for determining shot types and removing shadows. It is necessary to fill the fields.

Therefore, in the present invention, the upper and lower blocks of the current ground block are inspected using the conditions as shown in Equation 4 below to fill the holes as shown in FIG.

S300: shot boundary detection step

In the present invention, shot boundary detection is a basic process of video analysis. By classifying shots only at the shot boundary, the accuracy of shot type determination can be improved by reducing the calculation amount and reducing the determination error due to the temporary screen state change. In the present invention, for detecting the shot boundary, as shown in Equation 5 below, the temporal discontinuity (TBD) of the ground block map is checked to detect the shot boundary efficiently and quickly.

는 XOR 연산을 의미하며, i번째 영상프레임과 i-3 번째 영상프레임의 각 그라운드 블록(Ground Block)사이의 XOR 연산 결과의 총 합을 저장한다. 본 발명에서는 TBD_i-1 ＜ θ_SB 이고, TBD_i ＞ θ_SB 일 때, i번째 영상프레임을 샷 경계로 결정하였다. 또한, 샷 경계를 결정하는 임계값 θ_SB 은 30으로 설정하였다.here,

Denotes an XOR operation, and stores the total sum of the results of the XOR operation between each ground block of the i th video frame and the i-3 th video frame. In the present invention, when TBD _i-1 < θ _SB and TBD _i > θ _SB , the i-th image frame is determined as the shot boundary. In addition, the threshold value θ _SB for determining the shot boundary was set to 30.

S400: Shot type determination step

If a shot boundary is detected, shot type determination should be performed at that point. That is, the current shot is classified into a long shot and a non-long shot. In the case of long shots, a large number of ground blocks exist in the entire video frame. The shot classification is performed by measuring the length of the ground column of the center region of the soccer video using the measurement.

Find the longest green segment (LGS) for each column in the ground block map. If there is any LGS smaller than 1/3 of the height of the image within the horizontal golden area of the video frame (the center area equal to 5 when the image is divided into 3: 5: 3 horizontally), the shot will be It is determined as a non-long shot. That is, when defining the k-th image frame in the entire video, it can be expressed as Equation 6 below.

Here, θ _L = frame height / 3, GSLeft = frame width × 3/11, and GSRight = frame width × 8/11. In the present invention, the shot type determination is performed immediately after the shot boundary is detected, and the shot type determination is performed again after 3 frames and after 60 frames after the shot type is determined in preparation for an error in the shot type determination. This is in preparation for gradual shot conversion as well as shot determination error.

S500-S700: Analysis and Extraction Step of Ground Shadow Components

When a set of pixels in a height × width frame is R, R = {(x, y); 0 <x <height, 0 <y <width}. After the shadow region is extracted based on the characteristics of the analyzed shadow components, it is classified into the ground where the shadow exists, the ground without the shadow, and the non-ground area, and are represented as G _S , G _NS , and NG, respectively. That is, if P is defined as a divided form of R, it can be expressed as P = {G _S , G _NS , NG}.

In the present invention, since the image is limited to the football image, it is easy to separate the shadow area on the ground, and the YUV (Y: luminance signal, U: difference between the luminance signal and the red component, V :) of the image frame obtained directly at the output of the decoder. The shadow area on the ground is extracted using the difference between the luminance signal and the blue component) color space.

FIG. 5 is a diagram illustrating a state of each channel of a ground image in which a shadow region exists in the present invention, (a) an original image, (b) Y channel, (c) U channel, and (d) V channel, and FIG. Is a histogram for each channel of the area detected as ground in the image in which the shadow exists in the present invention, (a) Y channel, (b) U channel, Figure 7 is a shadow present in the present invention The figure shows the image and the shadow area on the extracted ground.

5 to 7, the shadow area has a lower brightness value than the bright area, and thus shows a relatively low Y value. And, in contrast to Y, it can be seen that the shadow area on the ground has a higher U value than the bright area, and there is no difference between the shadow area and the bright area for the V value.

In the present invention, the shadow region is extracted by using Y and U values showing a significant difference between the shadow region and the bright region.

That is, the histograms of the Y channel and the U channel are shown as shown in FIG. 6 only for the determined ground pixels using Equation 1 above. The horizontal axis represents bins of Y and U values between 0 and 255, and the vertical axis represents the cumulative number of pixels having the corresponding bin values.

As shown in the histogram, in the histograms of the Y channel and the U channel, each peak is present in the shadow area and the bright area, and thus, two peaks are present.

Among the two peaks, the peak of the shadow area is located toward the side having the lower index value, and for the U channel, the peak having the high index value corresponds to the shadow area. It becomes peak. As shown in Equation 7 below, the shadowed ground and the shadowless ground are distinguished by the positions of the peaks and the distance between the peaks.

Here, YPeak _Bright refers to the index value of the peak of the Y histogram corresponding to the bright ground area, and YPeak _Shadow refers to the index of the peak in the Y histogram corresponding to the shadowed ground area. (Index) value.

That is, the difference between the indexes of the two peaks is calculated, and if the difference is 50 or more, it is determined that daylight conditions are not constant in the ground and classified as candidate image frames having shadow areas.

Subsequently, in the present invention, a shadow region is extracted from the ground using a ground detection algorithm and a feature of a pixel in the simple shadow region in a candidate image frame in which the shadow region is expected to exist in the classified ground. can do.

Here, (x, y) means the current pixel, U (x, y) means the U value of the current pixel, and Y (x, y) means the Y value of the current pixel. In the soccer video, as shown in FIG. 5, the shadow area on the ground always shows that the U value of the pixel is higher than the Y value.

Using this property of the shadow area pixel, if the U value of the current pixel is larger than the Y value among the pixels determined as the ground pixel using Equation 1, the shadow area pixel is determined.

S800: shadow component reduction stage

In the present invention, in order to reduce the extracted shadow, first, the shadow area G _S is divided into two areas according to Equation 9 according to the intensity of the shadow component in the extracted shadow area.

Here, Y (x, y) means the Y value of the current pixel, and C _YPeaks is the peak index of the shadow region and the peak index of the bright region of the Y channel histogram obtained in FIG. It means the middle index of, and can be obtained in the same way as C _YPeaks = (YPeak _Bright + YPeak _Shadow ) / 2.

In the present invention, as described above, a pixel having a Y value corresponding to 75% or less of the C _YPeaks value is classified as a strong shadow ground area (G _SS ), and when it is 75% or more, a weak shadow ground area ( Weak Shadow Region, G _WS ). The shadow component may be reduced according to the intensity of the classified shadow region as shown in Equation 10 below.

In the present invention, YPeak _Diff and UPeak _{Diff, which} are the difference between the peak of the shadow areas of the green Y and U histograms and the peak index values of the bright areas, are used to increase the brightness of the shadow areas.

In addition, if the Y value of the pixel of the shadow area is all increased by the difference of the peak index, the existing bright area and the area of which the shadow is reduced are very unnaturally felt, and thus the brightness is brightened using the α and β values. You can get a natural image by adjusting.

At this time, the α and β values have a value in the range of [0, 1], and since the strong shadow area needs more brightness correction than the weak shadow area, the following relationship of 0 <β <α <1 Have

In addition, in order to obtain a more natural image, in the case of a strong shadow pixel, the difference between the Y value of the current pixel and the peak index of the shadow area is added, so that the Y value of the increased shadow area pixel is distributed more widely and evenly on the histogram. To help.

On the other hand, in the case of the U value, the U value of the shadow area is higher than that of the bright area, as opposed to Y. Therefore, the U value of the shadow pixel is lowered to be similar to that of the light area by subtracting UPeak _Diff from the U value of the current pixel.

8 is a view for comparing the results of performing the method of reducing the ground shadow in the athletic video according to an embodiment of the present invention, (a) Y histogram of the original image, (b) Y histogram of the result image .

As shown in (b) of FIG. 8, as a result of processing the shadow reduction, it can be seen that pixels corresponding to the shadow area that existed in the low area move to the right and are located in the bright area.

In order to perform the method of reducing the ground shadow in the athletic video constructed as described above, for example, it was developed using Visual Studio 2003 in a Win32 environment, and the ffmpeg library was used for MPEG decoding.

For the test, a soccer video with four shadows, Test 1, 2, 3, and 4, was used. Each video consists of 5000 frames, has 320 × 240 resolution, and is compressed using Xvid.

Ground map construction, shot boundary and type determination, and shadow reduction all run at around 32 fps or higher on a Pentium-4 3.0 GHz PC, enabling real-time operation on the PC.

Table 1 below shows the program execution speed when only the display is performed on the screen without any processing and when the shot boundary, the shot classification determination process, and the shadow reduction process are performed.

FIG. 9 is a view illustrating a process of processing a sports video according to the present invention, wherein (a) an original image in which shadow exists, (b) an image extracted from a shadow region, and (c) an image in which shadow is reduced. When the shadow component was decreased, the α value of Equation 10 was set to 0.8 and the β value was set to 0.4.

On the other hand, the method for reducing the ground shadow in the athletic video according to an embodiment of the present invention can also be implemented 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.

For example, the computer-readable recording medium may be a ROM, a RAM, a CD-ROM, a magnetic tape, a hard disk, a floppy disk, a removable storage device, a nonvolatile memory (Flash memory). Optical data storage, and the like, and also implemented in the form of a carrier wave (eg, transmission over the Internet).

The computer readable recording medium can also be distributed over computer systems connected over a computer network so that the computer readable code is stored and executed in a distributed fashion.

Although a preferred embodiment of the method for reducing the ground shadow in the athletic video according to the present invention has been described above, the present invention is not limited thereto, but the scope of the claims and the detailed description of the invention and the accompanying drawings are various. It is possible to carry out the transformation to this also belongs to the present invention.

According to the method of reducing the shadow of the ground in the athletic video of the present invention as described above, the shadow region is found in real time through a quick and simple method in the fast-moving image as in the field sports. By performing the shadow reduction for the user, there is an advantage that can provide a better viewing experience to the user.

Claims (13)

As a method of reducing ground shadows to improve the viewing experience of athletic videos, (a) determining whether there is a ground using an RGB value in each image frame of the athletic video; (b) constructing a ground block map by dividing an image frame having the ground into n × n blocks; (c) detecting shot boundaries by checking temporal discontinuity of the constructed ground block map; (d) determining a distant or non-distant shot at the detected shot boundary; (e) if it is determined in the step (d) that the far shot is made, determining whether a shadow area exists in the corresponding image frame; (f) extracting the shadow area using the YU value in the corresponding image frame when the shadow area exists in the step (e); And (g) classifying the first and second shadow areas according to the intensity of the shadow components in the extracted shadow area, and adjusting the brightness according to the intensity of the classified first and second shadow areas to reduce the shadow components. A method for reducing ground shadows in athletics video comprising steps. The method of claim 1, wherein in step (a), The ground is the shadow shadow reduction method of the athletic video, characterized in that determined by the following equation (1). [Equation 1]

Here, r, g, and b are RGB values at (x, y) coordinates in the video frame, and are values within the range of 0 to 255. In the (x, y) coordinates, G (x, y), which indicates whether it is grounded, is a value of 1 if it belongs to ground, or 0 otherwise. When I (x, y) is called the intensity value of the (x, y) coordinate, and if the intensity is 200 or more according to the value of I (x, y) (CASEⅠ) to a very bright ground, If the intensity is 50 or more and less than 200 (CASE II), the ground is a medium brightness, and if the intensity is less than 50 (CASE III), it is determined to be a dark ground. The method of claim 1, wherein in step (b), The ground block map is constructed by the following equations (2) and (3). [Equation 2]

[Equation 3]

Here, B _ij is a set of pixels in one block, GB (i, j) is a ground block map, and T _Ground is 0.5. The method of claim 1, wherein after step (b), And filling the holes generated in the ground block map. 5. The method of claim 4, wherein the holes are filled by inspecting and filling blocks above and below a current ground block according to Equation 4 below. 6. [Equation 4]

Where GB (i, j) is the ground block map. The method of claim 1, wherein in step (c), The shot boundary is detected by Equation 5 below. [Equation 5]

here,

Is an XOR operation, and stores the total sum of the result of the XOR operation between the i th image frame and each ground block of the i-3 th image frame. When TBD _i-1 &lt; θ _SB and TBD _i &gt; θ _SB , the i-th image frame is determined as the shot boundary, and θ _SB is set to 30. The method of claim 1, wherein step (d) (d-1) finding a segment having the longest ground color in each column of the ground block map; And (d-2) comparing the length of the longest segment with the ground color existing in the center region of the image frame with a predetermined reference value to determine the distance shot or the non-distance shot. How to reduce ground shadows in video. The method of claim 1, wherein in step (e), The shadow area distinguishes the shadowed ground and the shadowless ground by using the positions of the peaks in the histograms of the Y channel and the U channel and the distance between the peaks. How to reduce ground shadows. The method of claim 8, wherein the shadow area is determined by Equation 7 below. [Equation 7]

Here, YPeak _Bright is the index value of the peak of the Y histogram corresponding to the bright ground area, and YPeak _Shadow is the index of the peak in the Y histogram corresponding to the shadowed ground area. As a value of), the difference between the indexes of the two peaks is calculated, and if the difference is 50 or more, it is determined that daylight conditions are not constant in the ground and classified as candidate image frames having shadow areas. The method of claim 1, wherein in step (f), The shadow area is extracted by the following equation (8), the ground shadow reduction method of the athletic video. [Equation 8]

Here, (x, y) is the current pixel, U (x, y) is the U value of the current pixel, and Y (x, y) is the Y value of the current pixel. When the U value of the current pixel is larger than the Y value, the pixel is determined to be a pixel of the shadow area. The method of claim 1, wherein in step (g), The extracted shadow component is classified into the first and second shadow areas according to Equation 9 below. [Equation 9]

Where G _S represents the shadow area, Y (x, y) is the Y value of the current pixel, and C _YPeaks is the middle of the peak index of the shadow area and the peak index of the bright area of the Y channel histogram. As an index, C _YPeaks = (YPeak _Bright + YPeak _Shadow ) / 2. The first and second shadow areas are the strong shadow ground area G _SS and the weak shadow ground area G _WS , respectively. The method of claim 1, wherein in step (g), The classified first and second shadow areas reduce the shadow component by Equation (10) below. [Equation 10]

Here, YPeak _Diff and UPeak _Diff are the difference between peak and shadow index values of the shadow areas of the green Y and U histograms, respectively, and α and β values are in the range of [0, 1]. . A computer-readable recording medium having recorded thereon a program capable of executing the method of any one of claims 1 to 12.

Images