KR101426298B1 - apparatus and method for compensating image for enhancing fog removing efficiency - Google Patents

Abstract

The present invention relates to an image correcting method and an apparatus thereof capable of: generating a dark channel map which indicates a fog depth information of each pixel from a video frame using a dark channel prior, and remarkably reducing a block phenomenon which loses an edge information while efficiently correcting fog by performing a moving window algorithm on the generated dark channel map; not only efficiently removing fog, but also increasing an image quality by removing noise while preserving the edge information from the dark channel map using a bilateral filter; and efficiently removing more fog by finally determining a lower value as a dark channel prior value of a corresponding pixel by comparing the brightness values of the same pixels from a converted local-based dark channel map, and a filtered pixel-based dark channel map after an optimizing unit converts a window-based dark channel map, to which the moving window algorithm is applied, into the local-based dark channel map.

Description

[0001] The present invention relates to an apparatus and method for compensating for image fog,

The present invention relates to an image correcting apparatus and method for enhancing a fog removal rate. More specifically, a dark channel map including dark depth information is generated using a dark channel prior to a dark channel map, The map is refined and the brightness values of the refined dark channel map are rearranged by using a predetermined moving window algorithm to reduce the block phenomenon due to the loss of edge information and to improve the fog removal rate A correction device and a method.

Compared to the indoor image, the outdoor image has a variable brightness depending on the weather and weather conditions, and the quality of the image is deteriorated due to the noise characteristic. Especially, fog absorbs and scatters light, and images captured on a fogged day become less distant from color images and edge information as the distance from the subject increases.

Accordingly, various studies have been made on an image correction algorithm and an apparatus for correcting noise, especially fog, from a video image in a bad weather state to a clean image.

Conventionally, image correction methods for removing fog include a method of using different polarizing lenses, a method of using two or more cameras, and a method of using noise-free images. However, these methods require only one image, , There is a limit that can not be applied when there is no additional information (image data, GPS information, etc. for all places).

In addition, the image correction method using the conventional Markov random field (MRF) model is applied to the MRF model by using belief propagation (BP) or graph cut algorithm to correct the image, The method has a disadvantage in that a halo is generated in the resultant image due to an increase in saturation.

Also, in the conventional Fattal method, the vector direction of a clean image within a predetermined section is obtained by measuring the reflectance through ICA (independent component analysis) using the same vector direction as the vector direction of the reflectance R, Has a disadvantage that the correction rate is lowered for dense fog.

Accordingly, an image correction method using a dark channel prior, which has a simple calculation process and a high correction rate, has been studied.

The image correction method using the dark channel prior method is advantageous in that there is a pixel having a brightness of '0' in a certain region of a fogless image and that the darkest pixel existing in a certain region in the fog- And the amount of fog is calculated by using the property that the image is restored, and the amount of use is gradually increased because the image is restored and the calculation process is simple and the correction rate is high.

A fog removing device disclosed in Korean Patent No. 10-1190286 (entitled " Device and method for eliminating fog of an image based on edge information and tone mapping "), generates a dark channel map, Based on the dark channel flier value corresponding to the pixels of the region, the fog brightness value of the attraction image and the transmission amount indicating the degree of darkness of the fog are calculated and the image is restored.

However, in the fog removal apparatus, the transmission map generation unit refines the transmission amount using a soft matting algorithm to solve the block phenomenon in which the edge information of the dark channel map does not match the edge information of the original image , There is a disadvantage that the calculation processing time is excessively delayed because the soft matting algorithm processes the arithmetic operation based on the matrix requiring high arithmetic processing.

FIG. 1 is an exemplary view for explaining the problem of the Korean Registered Patent No. 10-11902586 (name of the invention: an apparatus and method for removing fog on an image based on edge information and tone mapping).

1, when the dark channel map is generated from the moving image frame 205 to which the image analysis unit is input, the fog removing unit does not directly use the brightness values of the respective pixels, Since a dark channel frier, which is fog depth information of each pixel, is calculated as a minimum value among the brightness values of the pixels included in the set size area 220, a block phenomenon occurs in which the edge of the transmission map does not coincide with the edge of the original image .

2 is a block diagram illustrating an image fog removal system as disclosed in U.S. Patent Application No. US2013 / 0071043 Al.

The image fog removing system 100 of FIG. 2 includes a pixel-based dark channel map generating unit 102 for generating a pixel-based dark channel map including dark channel values of pixels, Based dark channel map generation unit 104 for generating a local area based dark channel map in which a channel value is replaced with a minimum brightness value among brightness values of pixels included in a local region of a predetermined size, A final dark channel map generator 106 for generating an optimized final dark channel map using the channel map and the local area dark channel map, a fog brightness value calculator 106 for calculating the fog brightness value by analyzing the final dark channel map A transmission map generating unit 110 for generating a transmission map including a transmission amount indicative of the degree of fog using a fog brightness value calculated by the fog brightness value calculating unit, And a correction unit 112 for correcting the image by analyzing the amount of transmission of the moon map generation unit 110.

The prior art (100) shows that the final dark channel map generator 105 uses the pixel-based dark channel map and the locally-based dark channel map to calculate the brightness value for the pixel X of the pixel-based dark channel map, It is possible to more efficiently perform the fog removal by generating the final dark channel map in which the lower brightness value of the pixel X is determined as the brightness value of the pixel X. Thus, .

However, in the conventional art 100, since the brightness value of each pixel of the final dark channel map may be an accurate brightness value for the corresponding pixel, it may be a brightness value of the local region centered on the pixel. Therefore, The problem of FIG. 1 which does not coincide with that of FIG.

In addition, since the transmission amount of the transmission map is a value calculated based on the local area, the conventional technique 100 does not describe a separate method and means for reducing the block phenomenon in which the edge information does not coincide with the original image, Is lowered.

SUMMARY OF THE INVENTION The present invention has been made to solve the above problems, and it is an object of the present invention to provide a method and apparatus for generating dark channel maps representing fog depth information of each pixel from a moving picture frame using a dark channel prior, And a moving window algorithm for determining a lowest value as a dark channel flier value of the corresponding pixel by comparing the channel values of the input image and the channel values of the input image, .

Another object of the present invention is to save the edge information from the pixel-based dark channel map by using a blilter filter when the pixel-based dark channel map is generated, and reduce the processing throughput and speed by removing noise, And to provide an image correction apparatus and method capable of enhancing quality.

Yet another object of the present invention is to provide a method and an apparatus for optimizing an image processing apparatus in which an optimization unit converts a window-based dark channel map on which a moving window algorithm is performed to a local-based dark channel map, The present invention is to provide an image correcting apparatus and method capable of further increasing the fog removal efficiency by comparing the brightness values of the pixels and finally determining the lower value as the dark channel flier value of the corresponding pixel.

According to an aspect of the present invention, there is provided an image correction apparatus for removing fog from a moving image, the apparatus comprising: a pixel-based dark channel map including dark channel flier values representing fog depth information of each pixel from a current frame; Based dark channel map generating unit; Based on the pixel-based dark channel map generated by the pixel-based dark channel map generator, a combined average value of the Gaussian average value based on the brightness information and the Gaussian average value based on the distance, a filtering unit for performing a filer; A moving window algorithm that compares channel values for the same pixel from a predetermined number of previous frames from the current frame and determines a minimum value as a dark channel flier value of the corresponding pixel based on the pixel-based dark channel map filtered by the filtering unit; A moving window processing unit for generating a window-based dark channel map by executing the window-based dark channel map; A fog brightness value detector for detecting a fog brightness value of the current frame; A transmission map generating unit for generating a transmission map including a fog brightness value and a fog amount indicating a degree of fog of each pixel using the window based dark channel map; And an image correction unit for generating a reconstructed image from which the fog is removed from the current frame by using the fog brightness value and the transmission map.

delete

Also, in the present invention, a local-based dark channel map for replacing the brightness value of a specific pixel from the window-based dark channel map generated by the moving window processing unit with the lowest brightness value among local areas of predetermined sizes based on the specific pixel And compares the brightness values of the same pixel from the local-based dark channel map and the pixel-based dark channel map to finally determine the lower value of the two values as the brightness value of the corresponding pixel to generate an optimal dark channel map Preferably, the transmission map generating unit generates the transmission map using the optimal dark channel map generated by the optimizing unit.

Also, in the present invention, it is preferable that the fog brightness value detector determines the maximum brightness value excluding the upper 10% as the fog brightness value after generating the brightness value histogram from the current frame.

According to another aspect of the present invention, there is provided an image correction method for removing fog from a moving image, the method comprising: a fog brightness value detection step of detecting a brightness value from a current frame; A pixel-based dark channel map generating step of generating a pixel-based dark channel map consisting of dark channel flier values representing fog depth information of each pixel of the current frame; Based on the pixel-based dark channel map generated by the pixel-based dark channel map generating step, a combined average value of the Gaussian average value based on the brightness information and the Gaussian average value based on the distance, a filtering step of performing a filer; A moving window algorithm that compares channel values for the same pixel from a predetermined number of previous frames from the current frame and determines the lowest value as the brightness value of the pixel, based on the pixel-based dark channel map filtered by the filtering step; A moving window processing step of generating a window-based dark channel map by performing the following steps; Based on the window-based dark channel map generated by the moving window processing step and a fog brightness value detected by the fog brightness value detecting step, a transmission map including a delivery amount indicating a degree of darkness of fog contained in each pixel Generating a transmission map; And generating a reconstructed image, which is a fog removed image from the current frame, using the transmission map generated by the transmission map generating step and the fog brightness value.

Further, in the present invention, the Gaussian average value based on the brightness information from the pixel-based dark channel map generated by the pixel-based dark channel map generation step after the pixel-based dark channel map generation step, Further comprising a filtering step of performing a blilateral filer to determine a brightness value of each pixel with a combined average value of the Gaussian averages, wherein the moving window processing step includes a moving window algorithm . ≪ / RTI >

In the present invention, the brightness value of a specific pixel may be calculated from the window-based dark channel map generated by the moving window processing step after the moving window processing step, Based dark channel map is replaced with a lower brightness value, and then a brightness value of the same pixel is compared with the brightness value of the corresponding pixel from the local-based dark channel map and the pixel-based dark channel map, Further comprising an optimization step of finally determining an optimal dark channel map to generate an optimal dark channel map, and the transmission map generating step generates the transmission map using the optimum dark channel map generated by the optimization step.

Also, in the present invention, it is preferable that the fog brightness value detection step determines the maximum brightness value excluding the upper 10% as the fog brightness value after generating the brightness value histogram from the current frame.

According to the present invention having the above-mentioned solution, a dark channel map representing fog depth information of each pixel is generated from a moving picture frame using a dark channel prior, and a moving window algorithm is performed on the generated dark channel map, It is possible to remarkably reduce the block phenomenon in which the edge information is lost.

According to the present invention, edge information is preserved from a dark channel map using a blilter filter, but noise can be removed, thereby effectively removing fog and increasing image quality.

According to the present invention, the optimizing unit converts the window-based dark channel map, in which the moving window algorithm is performed, into the local-based dark channel map, and then converts the converted local-based dark channel map and the brightness value And finally determining the lower value as the dark channel flier value of the corresponding pixel, the fog removal is performed more efficiently.

FIG. 1 is an exemplary view for explaining the problem of the Korean Registered Patent No. 10-11902586 (name of the invention: an apparatus and method for removing fog on an image based on edge information and tone mapping).
2 is a block diagram illustrating an image fog removal system as disclosed in U.S. Patent Application No. US2013 / 0071043 Al.
3 is a block diagram illustrating an image correction apparatus according to an embodiment of the present invention.
FIG. 4 is a block diagram showing a moving window processing unit of FIG. 3. FIG.
Fig. 5 is an exemplary view of the figure.
FIG. 6 is a block diagram illustrating the dark channel map optimizer of FIG. 3. FIG.
Fig. 7 is an exemplary view of Fig. 6. Fig.
8 (a) is an actual screen showing an image to which the present invention is not applied, and FIG. 8 (b) is an actual screen showing a fog removed image according to the present invention.
FIG. 9 is a flowchart illustrating an operation procedure of the present invention.

Hereinafter, an embodiment of the present invention will be described with reference to the accompanying drawings.

3 is a block diagram illustrating an image correction apparatus according to an embodiment of the present invention.

The image correction apparatus 1 shown in Fig. 3 includes a database 12 in which data is stored, a communication interface 13 in which an image is received from the camera 10, Based dark channel map generator 14 for generating a dark channel map (hereinafter, referred to as a pixel-based dark channel map) consisting of a dark channel prior value, and a pixel- Based dark channel map filtered by the filtering unit 15 using the moving window algorithm, and a dark channel map of each pixel from the dark channel map filtered by the filtering unit 15 using the moving window algorithm, A moving window processing unit 16 for generating a window-based dark channel map by rearranging the flier values, and a moving-window processing unit 16 for comparing the pixel-based dark channel map and the window- A dark channel map optimizer 17 for generating a dark channel map (hereinafter, referred to as an optimal dark channel map) composed of optimal values of each pixel, a pixel-based dark channel map generator 14 for generating a dark- A fog brightness value detector 18 for detecting the fog brightness value by analyzing the dark channel map, and a transmission map for generating a transmission map indicating the fog level of each pixel using the fog brightness value and the optimum dark channel map An image correction unit 20 for analyzing the transmission map, the fog brightness value and the optimal dark channel map to correct the image, and the control objects 12, 13, 14, 15 ), (16), (17), (18), (19), and (20).

The control unit 11 is an OS (Operating System) of the image correction apparatus 1. The control unit 11 is an OS (Operating System) of the image correction apparatus 1 and includes control objects 12, 13, 14, 15, 16, 17, 19), (20).

The control unit 11 periodically crawls the communication interface unit 13 in real time and inputs the image received from the camera 10 to the pixel based dark channel map generating unit 14. [

The control unit 11 also inputs the pixel-based dark channel map generated by the pixel-based dark channel map generating unit 14 to the moving window processing unit 16, the optimizing unit 17 and the fog brightness value detecting unit 18.

The control unit 11 also inputs the window-based dark channel map generated by the moving window processing unit 16 to the optimizing unit 17.

The control unit 11 also receives an optimum dark channel map which is a dark channel map optimized by the dark channel map optimizing unit 17, a fog brightness value detected by the fog brightness value detecting unit 18, To the image correction unit 20,

The input video frame and the frame corrected by the video correction unit 20 (hereinafter referred to as a correction frame) are stored in the database unit 12.

The database unit 12 stores a pixel-based dark channel map generated by the pixel-based dark channel map generating unit 14, a window-based dark channel map generated by the moving window processing unit 16, An optimum dark channel map optimized by the fog brightness value detector 17, a fog brightness value detected by the fog brightness value detector 18, and a transmission map generated by the transmission map generator 19 are stored.

In addition, the database unit 12 stores the blilateral filers of FIG. 5, which will be described later, used in the filtering unit 15, and a moving window algorithm used in the moving window processing unit 16.

Hereinafter, an operation process for generating a dark channel map will be described.

The optical model operation of the following equation (1) is a well-known mathematical expression widely used for representing a fog image, and is defined for each of the R, G, and B channels.

[Equation 1]

는 안개가 포함된 카메라를 통해 획득된 영상인 입력영상이고,

는 복원해야할 깨끗한 영상이고,

는 영상의 안개 밝기값이고,

는 깨끗한 영상에 안개 밝기값이 어느 정도의 비율인지를 나타내는 안개의 짙은 정도로 정의되는 전달량이다.At this time

Is an input image which is a video obtained through a camera including a fog,

Is a clean image to be restored,

Is the fog brightness value of the image,

Is the amount of transmission defined as the degree of darkness of the fog indicating how much the fog brightness value is in a clean image.

로부터 안개 밝기값'

' 및 전달량 '

'를 산출하고, 이를 이용하여 최종적으로

을 검출하여 영상을 복원하도록 구성된다.Referring to Equation 1,

The fog brightness value "

'And delivery volume'

', And finally,

And reconstruct the image.

The pixel-based dark channel map generation unit 14 uses a point having a value adjacent to '0' in one of the pixels (one of the R, G, and B channels) within a predetermined interval in a clear image without fog A dark channel flier value of each pixel is calculated, and then a pixel-based dark channel map including a dark channel flier value of each calculated pixel is generated. At this time, the dark channel flier is defined by the following equation (2).

&Quot; (2) "

는 'x' 화소의 다크 채널 프라이어 값이다.At this time

Is the dark channel flier value of the 'x' pixel.

The pixel-based dark channel map generating unit 14 determines the lowest brightness value among the R, G, and B channels of each pixel as a dark channel flier, as shown in Equation (2). At this time, the dark channel flier value increases as the fog density becomes higher, and in the case of a fog free image, the dark channel flier value becomes '0' in most areas except sky.

The pixel-based dark channel map generated by the pixel-based dark channel map generating unit 14 is also supplied to the filtering unit 15, the dark channel map optimizing unit 17 and the fog brightness value detecting unit 18 ).

The filtering unit 15 preserves the edge information from the dark channel map input using the blilateral filer stored in the database unit 12 and removes the noise (fog), so that in the conventional image correction using the dark channel flier, Thereby reducing the block phenomenon. At this time, in the present invention, the filtering unit 15 uses a blilateral filer to process arithmetic based on a large matrix and to improve the conventional matting algorithm with high computation speed and high memory usage.

The blilateral filer is a nonlinear filter which combines a Gaussian filter using brightness information with a conventional Gaussian filter. In detail, the Gaussian average based on brightness information and the Gaussian mean value based on distance, The edge information of the image is preserved and the spot type fog and noise can be removed by filtering the brightness value of each pixel with a combined average value of the pixels. At this time, the blilateral filer is defined by the following equation (3).

&Quot; (3) "

는 필터링 된 이미지이고,

는 입력된 영상이고,

는 x를 중심으로 하는 기 설정된 영역인 윈도우이고,

은 밝기 차이를 평활화하기 위한 가우시안 필터이고,

는 거리 차이를 평활화하기 위한 가우시안 필터이다.At this time

Is a filtered image,

Is an input image,

Is a window which is a predetermined area centered at x,

Is a Gaussian filter for smoothing the brightness difference,

Is a Gaussian filter for smoothing the distance difference.

In this way, the blilateral filer is basically a single image filter. When two Gaussian windows are used in the filter, each of the two images is used to filter the edge information of the image, and the spot type fog and noise The computational throughput and speed can be reduced as compared with the conventional matting algorithm.

At this time, the filtered pixel-based dark channel map is input to the moving window processing unit 16 under the control of the control unit 11.

Fig. 4 is a block diagram showing a moving window processing unit of Fig. 3, and Fig. 5 is an illustration of the figure.

4 and 5, the dark channel map is conventionally determined to be the lowest brightness value in the local area of a predetermined size based on each pixel, thereby causing a block phenomenon in which the input image and the edge information do not match each other This is a solution to the problem.

In addition, the moving window processing unit 16 compares the dark channel flier value filtered by the filtering unit 15 with the first previous frame of the frame, and a comparison module for comparing the channel values of the same pixel from the second previous frame of the frame. A determination module 163 for determining the lowest channel value among the channel values compared by the comparison module 161 as a dark channel flier value of the corresponding pixel, And a window-based dark channel map generation module 165 for generating a window-based dark channel map consisting of the dark channel flier values of the window-based dark channel map.

At this time, the moving window algorithm applied to the moving window processing unit 16 is defined by the following equation (4).

&Quot; (4) "

That is, as shown in FIG. 5, the moving window processing unit 16 processes the dark channel flier values of the pixels to the same pixels of the input current frame 2 and the previous frames 2-1 and 2-2, It is possible to solve the conventional problem that the input image and the edge information do not coincide with each other by substituting the brightness value of the adjacent region.

At this time, the window-based dark channel map generated by the moving window processing unit 16 is input to the dark channel map optimizing unit 17 under the control of the control unit 11. [

FIG. 6 is a block diagram showing the dark channel map optimizer of FIG. 3, and FIG. 7 is an exemplary diagram of FIG. 6. FIG.

The dark channel map optimizer 17 determines an optimized dark channel flier value of each pixel using the input filtered pixel based dark channel map and the window based dark channel map.

The dark channel map optimizer 17 of FIG. 6 includes a conversion module 171 for converting a window-based dark channel map to a local-based dark channel map, a local-based dark channel map generated by the conversion module 171, A comparison module 173 for comparing brightness values of the same pixel from the dark channel map, a determination module 175 for determining a low brightness value of the two data by the comparison module 173 as a brightness value of the corresponding pixel, And an optimal dark channel map generation module 177 for generating an optimal dark channel map including the brightness values of the pixels determined by the determination module 175.

The conversion module 171 replaces the brightness value for a specific pixel from the window-based dark channel map generated by the moving window processing unit 16 with the lowest brightness value among local areas of a predetermined size based on a specific pixel And generates a locally-based dark channel map.

The comparison module 173 compares the brightness values for the same pixel from the local-based dark-channel map converted by the conversion module 171 and the pixel-based dark-channel map.

The determination module 173 determines a low brightness value as a brightness value of the corresponding pixel by the comparison module 173 and the optimal dark channel map generation module 177 determines an optimal dark pixel value having a brightness value determined by the determination module 173. [ Create a channel map.

That is, the dark channel map optimizer 17 converts the window-based dark channel map back to the local area, and takes a low value among the brightness values of the converted local area dark channel map and the pixel-based dark channel map, Can be further increased.

The optimal dark channel flier function of each pixel calculated by the dark channel map optimizing unit 17 is defined by the following equation (5).

&Quot; (5) "

The fog brightness value detection unit 18 determines the maximum brightness value excluding the upper 10% from the brightness value histogram generated after generating the brightness value histogram from the input image as the fog brightness value 'A', thereby selecting the brightest pixel from the input image Good results can be obtained. The fog brightness value function is defined by the following equation (6).

&Quot; (6) "

The transmission map generating unit 19 generates a transmission map based on the fog brightness value 'A' detected by the fog brightness value detector 18 and the darkness degree of the fog using the optimal dark channel map optimized by the dark channel map optimizer 17 , And generates a transmission map including the amount of transmission 't (x)' of each pixel. At this time, the transmission amount 't (x)' is defined by the following Equation (7).

&Quot; (7) "

The image correction unit 20 generates a reconstructed image from which fog is removed from the input image according to Equation (8).

&Quot; (8) "

는 안개가 제거된 복원영상이고,

는 기 설정된 전달량의 하한값이고,

는 안개 밝기값이다.At this time

Is a reconstructed image with fog removed,

Is a lower limit value of a predetermined amount of transmission,

Is the fog brightness value.

8 (a) is an actual screen showing an image to which the present invention is not applied, and FIG. 8 (b) is an actual screen showing a fog removed image according to the present invention.

8A and 8B, the moving window processing unit 16 compares the pixel values of the pixels with the pixel-based dark channel map and the brightness values of the same pixels from the previous frames, Value is determined, the block phenomenon in which the edge image is lost is remarkably reduced, so that the fog is efficiently removed, and at the same time, the edge information is not lost and the image is restored with high quality.

FIG. 9 is a flowchart illustrating an operation procedure of the present invention.

Referring to FIG. 9, the image correction method (S1) includes an image correction method using a dark channel flier (Dark channel prior), a pixel having a dark channel flier value, which is fog depth information of each pixel, Based dark channel map generated by the pixel-based dark channel map generation step (S10) using the previously stored blilateral filer is analyzed to generate a pixel-based dark channel map based on the pixel-based dark channel map A filtering step (S20) of preserving edge information from the dark channel map, removing noises, a current moving frame for the pixel-based dark channel map filtered by the filtering step (S20) using the previously stored moving window algorithm, Frame, and the second previous frames, and outputs the lowest channel value to the corresponding pixel A moving window processing step S30 for generating a window-based dark channel map by determining a brightness value, a window-based dark channel map generated by the moving window processing step S30 into a local-based dark channel map, An optimizing step (S40) of comparing the brightness values of the same pixel from the locally-based dark channel map and the pixel-based dark channel maps and finally determining a lower value among the two values as the brightness value of the corresponding pixel to generate an optimal dark channel map; A fog brightness value detection step (S50) of determining the highest brightness value as a fog brightness value by excluding the upper 10% of the generated histogram after generating a brightness value histogram from the input image, and a fog brightness value detection step (S50) The fog brightness value determined by the optimal dark channel map and fog brightness value detection step (S50) is used to calculate the delivery amount of each pixel, An image reconstruction step (S60) of generating a reconstruction image using the transmission map generated by the map generation step (S60), the transmission map generation step (S60), and the fog brightness value detected by the fog brightness value detection step (S50) S70).

1: image correction device 2: frame 3: specific pixel
11: control unit 12: database unit 13: communication interface unit
14: Pixel-based dark channel map generator 15:
16: Moving window processing unit 17: Dark channel map optimizing unit
18: fog brightness value detection unit 19: transmission map generation unit
20: image correction unit

Claims (8)

An image correction apparatus for removing fog from a moving image, comprising:
A pixel-based dark channel map generator for generating a pixel-based dark channel map consisting of dark channel flier values representing fog depth information of each pixel from the current frame;
Based on the pixel-based dark channel map generated by the pixel-based dark channel map generator, a combined average value of the Gaussian average value based on the brightness information and the Gaussian average value based on the distance, a filtering unit for performing a filer;
A moving window algorithm that compares channel values for the same pixel from a predetermined number of previous frames from the current frame and determines a minimum value as a dark channel flier value of the corresponding pixel based on the pixel-based dark channel map filtered by the filtering unit; A moving window processing unit for generating a window-based dark channel map by executing the window-based dark channel map;
A fog brightness value detector for detecting a fog brightness value of the current frame;
A transmission map generating unit for generating a transmission map including a fog brightness value and a fog amount indicating a degree of fog of each pixel using the window based dark channel map;
And an image correction unit for generating a reconstructed image from which the fog is removed from the current frame by using the fog brightness value and the transmission map. delete The method according to claim 1, further comprising the steps of: locally based on a specific pixel a brightness value of a specific pixel from a window-based dark channel map generated by the moving window processing unit, And compares the brightness values of the same pixel from the local-based dark channel map and the pixel-based dark channel map to finally determine the lower value of the two values as the brightness value of the corresponding pixel to generate an optimal dark channel map Further comprising an optimizer,
Wherein the transmission map generating unit generates the transmission map using the optimal dark channel map generated by the optimizing unit. The apparatus of claim 1 or 3, wherein the fog brightness value detector calculates a maximum brightness value excluding the upper 10% as a fog brightness value after generating a brightness value histogram from the current frame. 1. An image correction method for removing fog from a moving image, comprising:
A fog brightness value detection step of detecting a brightness value from the current frame;
A pixel-based dark channel map generating step of generating a pixel-based dark channel map consisting of dark channel flier values representing fog depth information of each pixel of the current frame;
Based on the pixel-based dark channel map generated by the pixel-based dark channel map generating step, a combined average value of the Gaussian average value based on the brightness information and the Gaussian average value based on the distance, a filtering step of performing a filer;
A moving window algorithm that compares channel values for the same pixel from a predetermined number of previous frames from the current frame and determines the lowest value as the brightness value of the pixel, based on the pixel-based dark channel map filtered by the filtering step; A moving window processing step of generating a window-based dark channel map by performing the following steps;
Based on the window-based dark channel map generated by the moving window processing step and a fog brightness value detected by the fog brightness value detecting step, a transmission map including a delivery amount indicating a degree of darkness of fog contained in each pixel Generating a transmission map;
And generating a reconstructed image that is a fog removed image from the current frame using the transmission map generated by the transmission map generating step and the fog brightness value. delete The method of claim 5, further comprising: after the moving window processing step, calculating a brightness value of a specific pixel from the window-based dark channel map generated by the moving window processing step, Based dark channel map is replaced with a lower brightness value, and then a brightness value of the same pixel is compared with the brightness value of the corresponding pixel from the local-based dark channel map and the pixel-based dark channel map, Further comprising an optimization step of finally determining and generating an optimal dark channel map,
Wherein the transmission map generation step uses the optimal dark channel map generated by the optimization step to generate a transmission map. The method of claim 5 or claim 7, wherein the fog brightness value detection step determines a maximum brightness value excluding the upper 10% as a fog brightness value after generating a brightness value histogram from the current frame.

Images