CN108810506A - A kind of Penetrating Fog enhancing image processing method and system based on FPGA - Google Patents

Abstract

The invention provides a method and system for image processing based on FPGA-based fog penetration enhancement, wherein the image processing system includes an FPGA chip and an ARM chip, the FPGA chip is connected to the ARM chip, an image data acquisition interface is provided on the FPGA chip, and the acquisition The image data is subjected to color shift correction processing, fog penetration enhancement processing, low contrast enhancement processing and/or detail enhancement processing; an output interface for outputting the processed image data is also provided on the FPGA chip. The present invention has processed the collected image for many times, so that the image quality has been greatly improved, and the image data collected under the condition of bad weather is not affected, and the effect of the finally obtained fog-free image is relatively good, which is different from the existing technology. Various monitoring equipment in the system can enhance and supplement the effect.

Description

A FPGA-based image processing method and system for enhanced fog penetration

technical field

The invention belongs to the technical field of image processing, in particular to an FPGA-based image processing method and system for enhanced fog penetration.

Background technique

After long-term development, video surveillance technology has now entered a relatively mature application stage. However, in some fields, it still cannot meet the specific observational requirements. For example, in environments such as smoke, fog, haze, water vapor, rain, snow, dust, darkness, and underwater, traditional monitoring equipment is difficult to function, or even helpless. In the era of extensive growth of the video surveillance market, this problem can be temporarily ignored. However, as manufacturers and users have a deeper understanding of the performance of technological equipment, the requirements have also increased accordingly.

In general, the existing video surveillance system often causes image quality degradation after image transmission and conversion, such as imaging, amplitude, scanning, transmission, and display lights. Due to insufficient or excessive lighting conditions during photography, the image will be too dark or too bright; distortion of the optical system, relative motion, atmospheric turbulence, etc. will make the image blurred, and various types of noise will also be introduced during the transmission process. In short, the input image may have many problems in terms of visual effect and recognition convenience. In order to solve the above problems, a Chinese patent with the publication number "CN105023256A" and the title "An Image Dehazing Method and System" was proposed. The image dehazing method of this patent needs to obtain the global atmospheric light and transmission of each channel of the image. The haze-free image of each channel is restored according to the global atmospheric light and transmittance of each channel, so as to obtain a haze-free image. However, the method of this patent is relatively complicated, and the effect of fog-free image processing is not good enough, and the obtained image quality is not high.

Contents of the invention

The object of the present invention is to provide an FPGA-based image processing method and system for fog penetration enhancement, which is used to solve the problem of poor image defogging processing effect in the prior art.

In order to achieve the above object, the present invention provides a FPGA-based fog penetration enhancement image processing system, including an FPGA chip and an ARM chip, the FPGA chip is connected to the ARM chip, and the FPGA chip is provided with an image data acquisition interface , and perform color shift correction processing, fog penetration enhancement processing, low contrast enhancement processing and/or detail enhancement processing on the collected image data; an output interface for outputting the processed image data is also set on the FPGA chip.

In order to solve the problem that the FPGA cannot be flexibly controlled, the ARM chip is provided with a network port for communicating with the host computer, and the ARM chip is used to control the FPGA to process the collected image data according to the instructions issued by the host computer.

Further, a gradient statistics module is also included, and the gradient statistics module is used for performing gradient statistics processing on the collected image data, and sending the image data after gradient statistics to the upper computer through the ARM chip.

The FPGA chip is provided with a first driver interface and a first debugging interface.

The ARM chip is provided with a second driver interface and a second debugging interface.

In order to monitor and observe the image conveniently, it also includes an electronic zoom module and a character superimposition module. After detail enhancement processing, the collected image data is output through the zoom processing of the electronic zoom module and the superposition processing of the character superposition module.

The present invention also provides a method for image processing based on FPGA-based fog penetration enhancement, comprising the following steps:

Collecting image data, performing color shift correction processing, fog penetration enhancement processing, low contrast enhancement processing and/or detail enhancement processing on the image data, and outputting the processed image data.

Further, gradient statistical processing is performed on the collected image data.

Further, the process of color shift correction processing is as follows: acquire the data of the R, G, and B channels of the collected image, establish a greenish intensity map representing the image, and select the pixel point with the smallest green component in the image as the overall image The measure of the degree of color cast refers to the corresponding value of the intensity map according to the set ratio in the intensity map as a quantitative representation of the color cast of the image.

In order to facilitate the monitoring and observation of the image, the collected image data is output through electronic zooming and character superposition processing after detail enhancement processing.

The beneficial effects of the present invention are:

The present invention includes an FPGA chip and an ARM chip, the FPGA chip is connected to the ARM chip, the FPGA chip is provided with an image data acquisition interface, and the collected image data is subjected to color shift correction processing, fog penetration enhancement processing, low Contrast enhancement processing and/or detail enhancement processing; an output interface for outputting the processed image data is also set on the FPGA chip. The present invention has processed the collected image for many times, so that the image quality has been greatly improved, and the image data collected under the condition of bad weather is not affected, and the effect of the finally obtained fog-free image is relatively good, which is different from the existing technology. Various monitoring equipment in the system can enhance and supplement the effect.

Description of drawings

Fig. 1 is the structure schematic diagram of the FPGA-based electronic fog penetration enhanced image processing system of the present invention;

Fig. 2 is a schematic flow chart of an image optimization algorithm of the present invention;

Figure 3-1 is a schematic diagram of the three-segment linear transformation of the fog penetration algorithm;

Figure 3-2 is a schematic diagram of calculating the dynamic range of brightness using the histogram statistical method of the fog penetration algorithm;

Figure 4-1 is a schematic diagram of the update of the column histogram on the far right of the kernel in the low contrast enhancement algorithm;

Figure 4-2 is a schematic diagram of the update of the neutralization histogram in the low contrast enhancement algorithm;

Figure 5 is a schematic diagram of detail enhancement algorithm;

FIG. 6 is a schematic diagram of bilinear difference processing of the electronic zoom algorithm.

Detailed ways

The specific embodiment of the present invention will be further described below in conjunction with accompanying drawing:

From the perspective of practical application, fog penetration technology includes the penetration of dust, water vapor, and small obstacles (such as slight dirt and rain on the transparent cover, etc.). In these harsh environments, the image quality will be greatly reduced or even unable to collect images of surveillance targets. Therefore, it is necessary to use fog penetration technology for post-processing and optimization of images. Image optimization processing mainly uses various algorithms to re-correct blurred images and reproduce live images. With the development of chip technology, the image optimization system on the computer is made into a single operating program and solidified on the FPGA. This is the so-called digital fog penetration technology. Since the pure digital fog penetration will not change the color, it is also called Color through the fog.

The FPGA-based electronic fog penetration enhanced image processing system of this embodiment includes an FPGA chip and an ARM chip, the FPGA chip is connected to the ARM chip, the FPGA chip is provided with an image data acquisition interface, and performs color shift correction processing on the collected image data , fog penetration enhancement processing, low contrast enhancement processing and/or detail enhancement processing; an output interface for outputting the processed image data is also set on the FPGA chip.

The ARM chip is provided with a network port for communicating with the host computer, and the ARM chip is used to control the FPGA to process the collected image data according to the instructions issued by the host computer. The FPGA chip is provided with a first driver interface and a first debugging interface, and the ARM chip is provided with a second driver interface and a second debugging interface.

The FPGA-based fog penetration enhanced image processing system of this embodiment also includes a gradient statistics module, which is used to perform gradient statistics processing on the collected image data, and pass the image data after gradient statistics through the ARM The chip is sent to the upper computer; in order to demonstrate the diversity functions of the image processing system of this embodiment, it also includes an electronic zoom module and a character superimposition module. After the details are enhanced, the collected image data is processed by the electronic zoom module. And the superposition processing output of the character superposition module.

Using the above-mentioned FPGA-based electronic fog penetration enhanced image processing system for image processing, as shown in Figure 2, mainly performs color shift correction processing, fog penetration enhancement processing, low contrast enhancement processing and/or detail enhancement on the collected image data Processing and other multiple processing, so that the obtained image quality is very high, does not affect the analysis of the collected images in bad weather.

Specifically, as shown in Figure 1, the image processing system includes a hardware part and a software part, and the hardware part mainly includes an interface, a processing core, peripherals, and a power supply module. Among them, the interface includes 1 channel HD-SDI input, 2 channels HD-SDI output, 2 channels PAL output, 1 channel RS422 communication, power supply and the first debugging interface, the first debugging interface includes RS422 debugging interface and JTAG debugging interface, ARM chip The JTAG debugging interface is shared with the FPGA chip, so the second debugging interface of the ARM chip is the JTAG debugging interface. The processing core uses the FPGA chip embedded with ARM, considering resources and power consumption, chooses Cycione VSX series FPGA of Altera Company. Peripherals mainly include the second drive interface external to ARM, namely DDR3 drive interface, 1 channel TF card, 1 channel NorFlash, the first drive interface externally connected to FPGA, namely DDR3 drive interface, and conversion chips for several interfaces. The power module receives 5V input and converts it to the voltage required by each chip on the board, such as 3.3V, 2.5V, 1.8V, 1.1V, etc.

The software part includes ARM software and FPGA logic. ARM loads the Linux operating system and implements basic drivers such as serial ports, DDR3 driver interfaces, and flash boot. The application layer is mainly responsible for communication command analysis, comprehensive control, and character mask drawing.

FPGA logic implements main image processing functions, including gradient statistics, color shift correction, fog penetration enhancement, low contrast enhancement, detail enhancement, electronic zoom, character overlay, image scaling, and interface frameworks such as interface driver and working mode configuration related information.

The image processing system is connected to a 1080p25/30 high-definition camera through the HD-SDI interface, and the 1080p color image data is input into the FPGA. After the data enters the FPGA, the main processing is realized through color shift correction, fog penetration enhancement, low contrast enhancement, and detail enhancement. Algorithm to complete the improvement of image quality and improve the observability; after electronic zoom and character superimposition, it will be output via HD-SDI interface (two channels of SDI output the same content); Scaling reduces the 1920*1080 image to 720*576, and then superimposes characters to output from the PAL interface. Here, the image data after detail enhancement processing is electronically zoomed and then image zoomed, because the electronic zoom processing can enlarge the image while keeping the output resolution unchanged, and the corresponding field of view will become smaller. If The data after detail enhancement processing is directly processed for image scaling, and the field of view does not change.

In order to realize the definition evaluation of the image, the image data input into the FPGA is processed by regional gradient statistics at the same time, and the statistical value is fed back to the ARM part, and transmitted to the host computer through the serial port.

On the one hand, ARM receives the data and commands transmitted by the host computer through the serial port, including the relevant data of superimposed characters and the control command of the working mode, and feeds back information such as self-test status and definition statistics to the host computer; on the other hand, it realizes The working mode is comprehensively controlled, and the drawing of the character mask is updated according to the obtained system information, and the character mask is provided to the character overlay module of the FPGA.

The image processing system takes the FPGA embedded with ARM as the processing core, configures related peripherals and interface circuits, and realizes real-time image processing. ARM software is mainly responsible for 3 parts:

(1) Integrated control of serial port communication: receive the mode control and system parameters from the host computer through the 422 serial port, issue instructions to the FPGA part according to the content of the mode control command, and control the switch of the FPGA working mode; at the same time, provide the system parameters to the character mask The graph drawing module is used as drawing input information; in addition, it completes functions such as system self-inspection and comprehensive task scheduling.

(2) Character mask drawing: Based on SKIA or other character drawing libraries, the character mask drawing is performed and cached to the DDR3 externally connected to the ARM for subsequent superimposition of the FPGA part.

(3) Interface driver and system boot: Complete system power-on configuration, including ARM part and FPGA part, complete the interface driver of ARM part, including DDR3, serial port, AXI bus, etc.

FPGA logic mainly includes 3 parts:

(1) Interface driver part: 1) Complete HD-SDI decoding and HD-SDI encoding, this part is completed by calling the IP provided by altera; 2) DDR3 cache control, mainly completes the input and output image cache, on the basis of the official provided IP, Perform multi-channel read-write control encapsulation; 3) SAF7129 driver, involving SAF7129I2C configuration and PAL data protocol conversion;

(2) Working mode control part: 1) Complete AXI bus communication with ARM, receive control commands from ARM, and return gradient statistics and other working status information; 2) Based on ARM control commands, configure each function module in the algorithm The working mode of the algorithm module.

(3) Algorithm function module: complete the processing of all algorithm function modules including gradient statistics, color shift correction, fog penetration enhancement, low contrast enhancement, detail enhancement, electronic zoom, image zoom, character superposition, and the interface between each module is unified Use Dval, DataRGB[23:0], CLK_pixel to realize full pipeline processing, and output 1 pixel processing result per pixel clock.

Each image optimization algorithm is described in detail below:

1) Color shift correction algorithm

Input: original image video sequence, video image resolution is 1920×1080 pixels, bit depth is 24bit.

Output: video image sequence after white balance, video image resolution is 1920×1080 pixels, bit depth is 24bit.

Realize process design:

Due to the different penetration capabilities of different bands of light in natural light, foggy images often have color cast problems when imaging. After the image is enhanced through fog, the color cast will also be enlarged and enhanced to become extremely obvious. Therefore, this algorithm is aimed at the sample video For the problem of greenish hue, a method of color cast correction is proposed before fog penetration is enhanced.

It is known that the R, G, and B three-channel pixel values of the input image I(x,y) are r(x,y), g(x,y) and b(x,y) respectively, and the greenish intensity map of the image is represented G(x,y), can be calculated by the following formula:

G(x,y)=max(0,g(x,y)-max(r(x,y),b(x,y)))

Among them, the max() operator means to calculate the maximum value of two numbers, and the value range of G(x,y) is [0,255]; the physical meaning of G(x,y) represents the green color contained in each pixel of the image how much. Considering that the characteristic of color cast images is that most of the images contain green components, in image processing, the pixel that contains the smallest green components in the whole image can be calculated as the measure of the overall color cast of the image, considering the single point selected In the actual process, the intensity map value G ₄₀ (x, y) corresponding to a certain proportion of points in the intensity map G(x, y) from small to large is selected as the quantitative representation of image color cast. The G channel pixel value g'(x,y) after color shift correction can be calculated by the following formula:

g'(x,y)=max(0,g(x,y)-G ₄₀ (x,y))

Compared with the traditional white balance algorithm, the color shift correction algorithm of this embodiment has the characteristics of small amount of calculation and strong adaptability, and the algorithm is not sensitive to the selection of the ratio value R _Thres . During the implementation process, the value range of R _Thres is selected [ 0.3,0.4] is better.

2) Fog penetration algorithm

Input: video image sequence after color shift correction, video image resolution is 1920×1080 pixels, bit depth is 24bit.

Output: the video image sequence after fog penetration, the video image resolution is 1920×1080 pixels, and the bit depth is 24bit.

Realize process design:

The fog penetration algorithm uses the global histogram stretching algorithm to realize the fog penetration enhancement function. Firstly, the histogram distribution of the foggy image is calculated, considering that the overall foggy image has a small number of histogram distributions in the low-brightness part, which causes the problem of low visual contrast. This algorithm considers the statistical distribution characteristics of the image pixel histogram, calculates the high and low brightness thresholds of image pixel stretching, and then increases the contrast through the R, G, and B three-channel histogram stretching method to achieve the effect of fog penetration.

Given that the R, G, and B three-channel pixel values of the input image I(x,y) are r(x,y), g(x,y) and b(x,y) respectively, then the grayscale image Y(x ,y) is calculated using the psychological formula of grayscale calculation:

Y(x,y)＝0.30*r(x,y)+0.59*g(x,y)+0.11*b(x,y)

During image enhancement processing, in order to highlight the target or gray-scale interval of interest, piecewise linear transformation can be used to divide the entire gray-scale interval into several gray-scale intervals, stretching the gray-scale interval corresponding to the target to be enhanced, which is relatively insensitive to suppression The gray level of interest, so as to achieve the purpose of enhancement. The commonly used piecewise linear transformation is three-segment linear transformation, as shown in Figure 3-1, and its mathematical expression is:

In the formula, M is the maximum brightness of the image. By adjusting the position of the inflection point of the broken line and the slope of the segmented line, that is, controlling the values of parameters a, b, c, and d, any gray-scale interval can be expanded or compressed.

The fog penetration algorithm uses the histogram of the statistical grayscale image Y(x, y), and calculates the dynamic range [LowThres, HighThres] of the image brightness by setting the low brightness ratio threshold Rlow and the high brightness ratio threshold Rhigh in the image, and finally determines three The parameters of segment linear transformation, the principle is shown in Figure 3-2. In the actual process, Rlow takes 0.005, Rhigh takes 0.01, a takes LowThres, b takes HighThres, c takes LowThres/3, and d takes (HighThres+255)/2.

3) Low contrast enhancement algorithm

Input: the video image sequence after fog penetration, the video image resolution is 1920×1080 pixels, and the bit depth is 24bit.

Output: low-contrast enhanced video image sequence, the video image resolution is 1920×1080 pixels, and the bit depth is 24bit.

Implementation process design: The image defogging algorithm often reduces the overall brightness of the image, and the original image details are lost to a certain extent due to the further reduction of the original brightness in areas with relatively low brightness, which requires the use of low contrast Enhanced module for brightness compensation.

The core of the low-contrast algorithm design is the estimation of the low-contrast area. The algorithm adopts a certain-scale median filtering method for the grayscale image, ignores the image details, and extracts the average brightness of the area where the pixel is located as the reference weight of the pixel enhancement. , that is, the brightness guide map; then select a nonlinear transformation function (exponential function is selected in the actual program), and use the brightness guide map to stretch the brightness of the three channels of the image R, G, and B to achieve contrast enhancement. It is known that the RGB three-channel pixel values of the input image I(x,y) are r(x,y), g(x,y) and b(x,y) respectively, and the grayscale image Y(x,y) of this module ) is calculated by the following formula:

Y(x,y)=max(r(x,y),g(x,y),b(x,y))

Grayscale image median filtering adopts fast median filtering algorithm. First, for each column of images, a histogram is maintained for it (for 8-bit images, the histogram has 256 elements). During the entire processing process, these Histogram data must be maintained. Each column of the histogram accumulates the information of 2r+1 adjacent pixels in the vertical direction (where r is the median filter radius). Initially, these 2r+1 pixels are centered on each pixel in the first row. of. The histogram of the kernel is obtained by accumulating 2r+1 adjacent column histogram data. These histogram data are kept up-to-date with constant time in two steps throughout the filtering process. As shown in Figures 4-1 and 4-2, consider the case of moving one pixel to the right from a certain pixel. For the current row, the column histogram on the far right of the kernel needs to be updated first, and at this time the data in the column histogram of this column is still calculated centered on the pixel at the corresponding position of the previous row. Therefore, it is necessary to subtract the histogram corresponding to the uppermost pixel and then add the histogram information of the pixel below it. The effect of this is to lower the column histogram data by one line. This step is obviously a 0(1) operation, with only one addition and one subtraction, regardless of the radius r. The second step updates the kernel histogram, which is the sum of 2r+1 column histograms. This is obtained by subtracting the leftmost column histogram data and adding the column histogram data for the column processed in the first step. As mentioned earlier, the time-consuming addition, subtraction, and calculation of the median value of the histogram are all calculations that depend on the bit depth of the image, and have nothing to do with the filter radius.

Using the brightness guide map, set the three-channel gray value of the image R, G, and B to Y _r/g/b (x, y), and the three-channel gray value after gray scale stretching is Y' _r/g/b (x,y), the non-linear function used for grayscale stretching is:

Among them, Meduim(x, y) is the grayscale point, Y _r/g/b (x, y) corresponds to the pixel value of the median filter image, Y _min is the minimum pixel value of the grayscale image, and Y _avg is the grayscale image Pixel average.

4) Detail enhancement algorithm

Input: low-contrast enhanced video image sequence, the video image resolution is 1920×1080 pixels, and the bit depth is 24bit.

Output: video image sequence after detail enhancement, video image resolution is 1920×1080 pixels, bit depth is 24bit.

Implementation process design: the detail enhancement algorithm is an optional module in the embodiment, and its goal is to enhance image detail information. The important process of the algorithm adopts the classic Unsharp Mask (USM) sharpening algorithm. The specific principle is shown in Figure 5, and the specific formula is:

y(n,m)=x(n,m)+λz(n,m)

Among them, x(n,m) is the input image, y(n,m) is the output image, λ is a scaling factor used to control the enhancement effect, and z(n,m) is the correction signal, generally through the input image Perform high-pass filtering on x to obtain, in this algorithm:

z(n,m)=x(n,m)-g(n,m)

Among them, g(n,m) is the Gaussian filtering result of x(n,m).

In the process of image processing using the detail enhancement algorithm, the R, G, and B channels of the original image are first low-pass filtered, and then the difference between the original image and the low-pass filtered image is used to extract the high-frequency part of the image, and finally Weighted superposition of high-frequency images to achieve image detail enhancement.

5) Electronic zoom algorithm

Input: video image sequence after detail enhancement, the video image resolution is 1920×1080 pixels, and the bit depth is 24bit.

Output: video image sequence after 2x or 4x electronic zoom, video image resolution is 1920×1080 pixels, bit depth is 24bit.

Realize process design: The purpose of electronic zoom is to enlarge and display the central area of the image to meet the needs of monitoring and observation. It is divided into two functional requirements: 2× electronic zoom and 4× electronic zoom.

The design of the algorithm considers both efficiency and effect, and uses a bilinear interpolation algorithm to realize the electronic zoom function. Mathematically, bilinear interpolation is a linear interpolation extension of an interpolation function with two variables, and its core idea is to perform a linear interpolation in two directions, as shown in Figure 6.

If you want to get the value of the unknown function f at point P=(x, y), and the known function f is at Q ₁₁ = (x ₁ , y ₁ ), Q ₁₂ = (x ₁ , y ₂ ), Q ₂₁ = (x ₂ , y ₁ ) and Q ₂₂ = (x ₂ , y ₂ ) the values of the four points, the calculation and solution process is as follows:

First perform linear interpolation in the x direction to get:

Then perform linear interpolation in the y direction to get:

This results in an expression for f(x, y:

If a coordinate system is chosen so that the coordinates of the four known points of f are (0, 0), (0, 1), (1, 0) and (1, 1), then the interpolation formula can be simplified as:

f(x,y)≈f(0,0)(1-x)(1-y)+f(1,0)x(1-y)+f(0,1)(1-x)y+ f(1,1)xy.

Or expressed in matrix operations as:

Unlike this interpolation method, the result of this interpolation method is usually not linear, expressed as:

b ₁ +b ₂ x+b _3y +b ₄ xy.

where the number of constants corresponds to the number of data points for a given f:

b ₁ =f(0,0)

b ₂ =f(1,0)-f(0,0)

b ₃ =f(0,1)-f(0,0)

b ₄ =f(1,1)-f(1,0)-f(0,1)+f(0,0)

The result of linear interpolation is independent of the order of interpolation. The interpolation in the y direction is performed first, and then the interpolation in the x direction is performed, and the result obtained is the same.

6) Gradient statistics algorithm

Input: Original video image sequence, video image resolution is 192×108 pixels, bit depth is 24bit.

Output: gradient reference value.

Implementation process design: The gradient statistical algorithm mainly uses the principle of contrast focusing to realize automatic focusing by detecting the contour edge of the image. The sharper the edge of the outline of the image, the greater its brightness gradient, or the greater the contrast between the scene and the background at the edge. Conversely, for an out-of-focus image, the contour edge is blurred, and the brightness gradient or contrast decreases; the farther out of focus, the lower the contrast. Using this principle, when the absolute value of the output contrast difference is the smallest, it means that the focus is completed.

This algorithm uses the Sobel gradient operator as the measure of variance to calculate the gradient value of the image center area (192×108 pixels). When the gradient value reaches the maximum, the focus is completed. The Sobel operator is mainly used for edge detection. Technically, it is a discrete difference operator, which is used to calculate the approximate value of the gray level of the image brightness function. Using this operator at any point in the image will generate the corresponding grayscale vector or its normal vector. The Sobel convolution factor is:

The operator includes two sets of 3x3 matrices, which are horizontal and vertical respectively, and plane convolution is performed with the image to obtain the approximate values of the horizontal and vertical brightness differences respectively. If A represents the original image, Gx and Gy represent the gray value of the image detected by horizontal and vertical edges respectively.

The horizontal and vertical gray values of each pixel of the image are combined by the following formula to calculate the gray value of the point:

In order to improve efficiency, the gradient statistical algorithm uses an approximation that does not take the square root as the gradient value:

|G|＝|G _x |+|G _y |

7) Local definition statistics module

This module is to make statistics on the sharpness of each pixel in the image data, and is used to judge the enhancement effect of fog penetration, and the recording is completed by ARM cpu.

Specific implementations have been given above, but the present invention is not limited to the above-described implementations. The basic idea of the present invention lies in the above-mentioned basic scheme. For those of ordinary skill in the art, according to the teaching of the present invention, it does not need to spend creative labor to design various deformation models, formulas, and parameters. Changes, modifications, substitutions and deformations to the embodiments without departing from the principle and spirit of the present invention still fall within the protection scope of the present invention.

Claims (10)

1. a kind of Penetrating Fog based on FPGA enhances image processing system, which is characterized in that including fpga chip and ARM chips, institute It states fpga chip to connect with the ARM chips, image data acquiring interface is provided on the fpga chip, and to the figure of acquisition As data carry out color cast correction processing, Penetrating Fog enhancing processing, low contrast enhancing processing and/or details enhancing processing；It is described It is also set up on fpga chip for by the output interface of treated image data output.

2. the Penetrating Fog according to claim 1 based on FPGA enhances image processing system, which is characterized in that the ARM cores On piece is provided with the network interface for being communicated to connect with host computer, and the ARM chips are used to be controlled according to the instruction that host computer issues FPGA handles the image data of acquisition.

3. the Penetrating Fog according to claim 2 based on FPGA enhances image processing system, which is characterized in that further include gradient Statistical module, the gradient statistical module is used to carry out gradient statistical disposition to the image data of acquisition, and will pass through gradient and unite Image data after meter is sent to host computer by the ARM chips.

4. the Penetrating Fog according to claim 1 based on FPGA enhances image processing system, which is characterized in that the FPGA cores On piece is provided with the first driving interface and the first debugging interface.

5. the Penetrating Fog according to claim 4 based on FPGA enhances image processing system, which is characterized in that the ARM cores On piece is provided with the second driving interface and the second debugging interface.

6. the Penetrating Fog according to claim 1 based on FPGA enhances image processing system, which is characterized in that further include electronics Zoom module and character adding module, the image data of acquisition is after details enhancing processing, by the zoom of Electronic magnification module The output of the overlap-add procedure of processing and character adding module.

7. a kind of Penetrating Fog based on FPGA enhances image processing method, which is characterized in that include the following steps：

Image data is acquired, color cast correction processing, Penetrating Fog enhancing processing, low contrast enhancing processing are carried out to described image data And/or details enhancing processing, and image data exports by treated.

8. the Penetrating Fog according to claim 7 based on FPGA enhances image processing method, which is characterized in that the figure of acquisition As data have carried out gradient statistical disposition.

9. the Penetrating Fog according to claim 8 based on FPGA enhances image processing method, which is characterized in that the colour cast school The process just handled is：The data in tri- channels R, G, B of the image of acquisition are obtained, the characterization partially green intensity map of image, choosing are established Measurement of the pixel of the minimum containing green components in image as image entirety colour cast degree is taken, according to setting in the intensity map Fixed ratio refers to quantization means of the corresponding intensity map values as image color cast.

10. the Penetrating Fog according to claim 7 based on FPGA enhances image processing method, which is characterized in that the figure of acquisition Picture data are after details enhancing processing, by Electronic magnification and character adding processing output.