CN112053309A - Image enhancement method and image enhancement device - Google Patents

Abstract

An image enhancement method, comprising: obtaining a source image I_in(u, v); stretching I_inR, G and B-channel color amplitude of (u, v), resulting in color amplitude stretched image I_stre(u, v); conversion I_stre(u, v) color space of (I)_stre(u, V) converting from RGB space to HSV space to obtain hue component H (u, V), brightness component V (u, V) and saturation component S (u, V); weighting and calculating V (u, V) to obtain a new V channel image V_out(u, v); stretching S (u, v) to obtain a stretched S channel image S_out(u, v); inverse transformation of H (u, V), V_out(u, v) and S_out(u, v) to RGB space to obtain image I after image enhancement_out(u, v). The embodiment of the invention also provides a device for implementing the image enhancement method.

Description

Image enhancement method and image enhancement device

technical field

The present invention relates to the technical field of image processing, and in particular, to an image enhancement method and an image enhancement device.

Background technique

Color digital images are an important medium for feeding back information about nature. Under unfavorable shooting conditions, the captured images usually have low contrast and even degraded images with severe color casts.

To obtain more information from low-quality images, some algorithms for image enhancement are proposed in the prior art to optimize the image. Perform equalization operations on the sub-histograms respectively; the modified histogram equalization algorithm can control the effect of the algorithm by modifying the histogram frequency value and the cumulative distribution function; the local histogram equalization algorithm implements local equalization operations according to the spatial position of the image ; There is also image enhancement technology based on transform domain equalization, which is equalized image enhancement by transforming the image in the spatial domain to other domains. Although these methods play an optimization role, they also have drawbacks: the sub-histogram equalization algorithm needs to perform multiple equalization operations, and it is difficult to select an appropriate threshold; although the modified histogram equalization algorithm only needs to perform an equalization operation, It is difficult to find a satisfactory correction method and clipping parameters; the histogram variational specification technology algorithm has a large amount of calculation, and it is difficult to design a suitable target histogram; it is difficult to find a suitable method for the local histogram equalization algorithm "block effect" phenomenon and "over-enhancement" phenomenon; image enhancement technology based on transform domain equalization is the algorithm complexity is too high.

SUMMARY OF THE INVENTION

In view of this, in order to obtain more information from low-quality images and overcome the above drawbacks, it is necessary to provide an image enhancement method and an image enhancement apparatus.

An embodiment of the present invention provides an image enhancement method, comprising the following steps:

Obtain the source image I _in (u, v), I _in (u, v)={R _in (u, v), G _in (u, v), B _in (u, v)};

Stretch the color amplitudes of the R, G and B channels of I _in (u,v) to obtain the image I _stre (u, v) after the color amplitude is stretched, I _stre (u, v) = {R _stre (u, v ),G _stre (u,v),B _stre (u,v)};

Convert the color space of I _stre (u, v), so that I _stre (u, v) is converted from RGB space to HSV space, and obtains the hue component H(u, v), the luminance component V(u, v) and the saturation component S(u,v);

Weighted calculation V(u, v) to obtain a new V channel image V _out (u, v);

Stretch S(u, v) to obtain the stretched S channel image S _out (u, v);

Inversely transform H(u,v), _Vout (u,v) and _Sout (u,v) to RGB space to obtain the image Iout(u,v) after image enhancement, _Iout (u,v)={ _Rout (u,v), _Gout (u,v), _Bout (u,v)}.

Embodiments of the present invention further provide an image enhancement device, which may include:

an acquisition unit for acquiring a source image I _in (u, v), I _in (u, v)={R _in (u, v), G _in (u, v), B _in (u, v)};

The RGB stretching unit is used to stretch the R, G and B channel color amplitudes of I _in (u, v) to obtain the image I _stre (u, v) after the color amplitude is stretched, I _stre (u, v) = {R _stre (u,v),G _stre (u,v),B _stre (u,v)};

The conversion unit is used to convert the color space of I _stre (u, v), so that I _stre (u, v) is converted from the RGB space to the HSV space, and the hue component H(u, v) and the luminance component V(u, v are obtained. ) and the saturation component S(u,v);

a calculation unit for weighted calculation V(u, v) to obtain a new V channel image V _out (u, v);

The saturation component stretching unit is used to stretch S(u, v) to obtain the stretched S channel image S _out (u, v);

An inverse transform unit for inversely transforming H(u,v), _Vout (u,v) and _Sout (u,v) to RGB space to obtain an image-enhanced image _Iout (u,v), _Iout (u,v)={ _Rout (u,v), _Gout (u,v), _Bout (u,v)}.

In the embodiment of the present invention, the R, G, and B channel color amplitudes of the source image are first stretched, and then the V channel image and the S channel image are adjusted based on the HSV space, so as to equalize the histogram of the V channel image and maximize the gray value of the S channel image. Stretch, and then inversely transform the adjusted image to RGB space to obtain an enhanced image. The method provided by the embodiment of the present invention can overcome the "over-enhancement" phenomenon caused by the traditional histogram equalization algorithm, and the calculation process is simple and complicated. It can effectively improve the contrast and brightness information of the image, and obtain an enhanced image with high contrast and high color brightness.

Description of drawings

In order to explain the technical solutions of the embodiments of the present invention more clearly, the following will briefly introduce the drawings that need to be used in the embodiments. Obviously, the drawings in the following description are some embodiments of the embodiments of the present invention. For those of ordinary skill in the art, other drawings can also be obtained from these drawings without any creative effort.

FIG. 1 is a flow chart of steps of an image enhancement method according to a preferred embodiment.

FIG. 2 is a diagram showing the effect of an implementation method of an embodiment of the present invention.

FIG. 3 is a reference diagram of changes before and after the weighted histogram equalization algorithm is implemented on the V channel in an embodiment of the present invention.

FIG. 4 is a reference diagram of changes before and after the weighted equalization histogram is remapped according to an embodiment of the present invention.

FIG. 5 is a reference diagram of changes before and after the saturation maximization stretching algorithm is implemented on the S channel according to an embodiment of the present invention.

FIG. 6 is a comparison reference diagram of enhancement results of six schemes for a scene graph named "earth" according to an embodiment of the present invention.

FIG. 7 is a comparison reference diagram of histogram results of six schemes for a scene graph named "earth" according to an embodiment of the present invention.

FIG. 8 is a comparison reference diagram of enhancement results of six schemes for a scene graph named “road” according to an embodiment of the present invention.

FIG. 9 is a comparison reference diagram of histogram results of six schemes for a scene graph named “road” according to an embodiment of the present invention.

FIG. 10 is a comparison reference diagram of enhancement results of six schemes for a scene graph named “126007” according to an embodiment of the present invention.

FIG. 11 is a comparison reference diagram of histogram results of six schemes for a scene graph named “126007” according to an embodiment of the present invention.

FIG. 12 is a comparison reference diagram of enhancement results of six schemes for a scene graph named "5096" according to an embodiment of the present invention.

FIG. 13 is a reference diagram for comparing the histogram results of six schemes for a scene graph named “5096” according to an embodiment of the present invention.

FIG. 14 is a schematic diagram of the structure and composition of an image enhancement apparatus according to a preferred embodiment.

FIG. 15 is a schematic diagram of the structure and composition of the RGB stretching unit 12 according to the first preferred embodiment.

FIG. 16 is a schematic diagram of the structure and composition of the RGB stretching unit 12 according to the second preferred embodiment.

FIG. 17 is a schematic diagram of the structure and composition of the RGB stretching unit 12 according to the third preferred embodiment.

FIG. 18 is a schematic structural diagram of the computing unit 14 according to a preferred embodiment.

FIG. 19 is a schematic diagram of the structure and composition of the weighting calculation unit 141 in a preferred embodiment.

Detailed ways

The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

An embodiment of the present invention provides an image enhancement method, and the main body implementing the method is an image enhancement apparatus, and the apparatus may be a terminal device with an image processing function. Referring to FIG. 1 and FIG. 2 , the implementation steps of the image enhancement method provided by the embodiment of the present invention may specifically include:

Step S110, acquiring a source image I _in (u, v), where I _in (u, v)={R _in (u, v), G _in (u, v), B _in (u, v)};

Step S111, stretch the color amplitudes of the R, G and B channels of the source image I _in (u, v) to obtain an image I _stre (u, v) after the color amplitude is stretched, where I _stre (u, v)={R _stre (u,v),G _stre (u,v),B _stre (u,v)};

Step S112, convert the color space of the image I _stre (u, v), so that the image I _stre (u, v) is converted from the RGB space to the HSV space to obtain the hue component H(u, v), the luminance component V(u, v) ) and the saturation component S(u,v);

Step S113, weighted calculation V(u, v) to obtain a new V channel image V _out (u, v);

Step S114, stretch S(u, v) to obtain the stretched S channel image S _out (u, v);

Step S115, inversely transform H(u,v), _Vout (u,v) and _Sout (u,v) to RGB space to obtain an image enhanced image _Iout (u,v), _Iout (u, v)={ _Rout (u,v), _Gout (u,v), _Bout (u,v)}.

In the embodiment of the present invention, the input color source image I _in (u, v) in RGB format is firstly in RGB space, and its R, G and B channels are respectively subjected to color amplitude stretching, and the source image I _in (u, v ) can be expressed as:

I _in (u,v)={R _in (u,v),G _in (u,v),B _in (u,v)} (1)

where (u, v) represents the pixel position of the source image and satisfies u=1,...,U,v=1,...,V.

To reduce image distortion caused by unfavorable capture environments, RGB space color channel stretching can be used as a preprocessing algorithm for image contrast enhancement, stretching each color channel to the maximum allowed range. Stretch the color amplitudes of the R, G and B channels of I _in (u,v) to obtain the image I _stre (u,v) after the color amplitude is stretched. The image I _stre (u,v) can be expressed as:

I _stre (u,v)={R _stre (u,v),G _stre (u,v),B _stre (u,v)} (2)

Specifically, the step of realizing the R channel color amplitude of stretching I _in (u, v) may specifically include: calculating the value range of the R channel color amplitude of I _in (u, v); selecting the calculated R channel color amplitude to take The highest value in the value range; stretch the R channel color amplitude to the highest value in the R channel color amplitude value range to get R _stre (u,v).

The color stretching formula of the R channel can be expressed by formula (3)

where min({R(u,v)}) and max({R(u,v)}) represent the minimum and maximum values of pixels in the R channel, respectively.

Specifically, the step of realizing the G channel color amplitude of stretching I _in (u, v) may specifically include: calculating the value range of the G channel color amplitude of I _in (u, v); selecting the calculated G channel color amplitude to take The highest value in the value range; stretch the G channel color amplitude to the highest value in the value range of the G channel color amplitude to obtain G _stre (u,v).

The color stretching formula of G channel can be expressed by formula (4)

where min({G(u,v)}) and max({G(u,v)}) represent the minimum and maximum values of pixels in the R channel, respectively.

Specifically, the step of realizing the B channel color amplitude of stretching I _in (u, v) may specifically include: calculating the value range of the B channel color amplitude of I _in (u, v); selecting the calculated B channel color amplitude to take The highest value in the value range; stretch the B channel color amplitude to the highest value in the B channel color amplitude value range to get B _stre (u,v).

The color stretching formula of the B channel can be expressed by formula (5)

where min({B(u,v)}) and max({B(u,v)}) represent the minimum and maximum values of pixels in the R channel, respectively.

An image in RGB format has difficulty distinguishing the chrominance, brightness and saturation information of the image. If the contrast of the source image is directly enhanced in the RGB color space, the proportional relationship between R, G and B is easily destroyed, resulting in color distortion. The HSV (hue, saturation, value) color space is more in line with the human experience of color perception than other color spaces, and can greatly reduce the mutual interference between the detail information of the image and the color information. In step S112, the color space of _Isre (u, v) is converted, so that _Isre (u, v) is converted from the RGB space to the HSV space, and the hue component H(u, v) and the luminance component V(u, v) are obtained ) and the saturation component S(u,v). The conversion function for converting the RGB color space to the HSV color space can be expressed by formula (6):

The traditional histogram equalization algorithm usually produces artifacts due to "over-enhancement". The evenly distributed histogram can retain more detailed information of the image. The embodiment of the present invention combines the gray levels of the histogram through an optimization criterion. optimization, so as to obtain a satisfactory histogram and achieve the purpose of improving the image quality.

In step S113, a weighted histogram equalization operation is performed on the V channel, and V(u, v) is weighted to obtain a new V channel image V _out (u, v). The specific implementation may include the following steps: (u, v) perform weighted calculation corresponding to the gray value of the histogram h(i) to obtain the weighted histogram h _W (i); normalize h _W (i) to obtain h _N (i); map The non-null intensity levels in h _N (i) make the histogram distribution of h _N (i) uniform, resulting in a weighted calculated V _out (u,v).

The weighted histogram h _W (i) can be expressed by formula (7):

h _W (i)=β×h(i)+ω×max{(h(i))},0＜ω＜1,ω+β=1 (7)

Among them, h _W (i) is the weighted histogram, h(i) is the histogram, max{h(i)} is the maximum gray value, i=0,1,...,L-1, when h(i)=(L-1), then satisfy

h _W (i) _max =(ω+β)×(L-1)=(L-1) (8)

From formulas (7) and (8), it can be concluded that the optimization criterion can ensure that the output gray level is always kept in the range of [0, L-1]. For the darker part of the source image, the optimization criterion can increase the low gray value; for the brighter part of the source image, the optimization criterion makes the high gray value smaller. It can effectively avoid excessive enhancement of the histogram, reduce the generation of artifacts to a great extent, and ensure that the contrast of the image is enhanced while maintaining the brightness of the image. The changes before and after the weighting of the histogram are shown in FIG. 3 . It can be seen from Fig. 3b in Fig. 3 that the low gray value can recover more information content after the weighted equalization optimization.

The embodiment of the present invention normalizes h _W (i) to obtain h _N (i), wherein the histogram h _N (i) can be represented by formula (9):

Map the non-null intensity levels in _hN (i) to make the histogram distribution of _hN (i) uniform, and obtain the weighted _Vout (u,v).

The equalization operation of the weighted optimized histogram may still cause some loss of gray value. As shown in Figure 3, the lost gray value can be detected by observing the change of the histogram. Therefore, the non-empty gray value needs to be The grayscale values are remapped to the entire interval in order to obtain a histogram with a uniform distribution.

First specify a set Ω to store non-empty grayscale values. The formula is defined as follows:

Ω={Ω(m)=h _N (i)|h _N (i)>0} (10)

Wherein, Ω(m) will store the number of non-null grayscale values generated after histogram equalization, and satisfy m=1, 2, . . . , m _max . If m≠0, it needs to be remapped to [0,L-1]. Use the following calculation rules to ensure that the final output histogram covers the entire space evenly. The calculation guidelines are as follows:

The histogram remapping process is shown in Figure 4. Figure 4 clearly shows that before the histogram remapping, the histogram is obviously unevenly distributed. After remapping, all non-empty gray values are evenly distributed in [0, 1] range.

In addition, the embodiment of the present invention also provides a selection scheme for the weight parameters ω and β involved in formula (7), and searches and selects through the golden section algorithm: select the initial search range; select the brightness error as the iterative Objective function; iterate through the objective function to narrow the initial search range; repeat the iterative operation through the narrowed initial search range until the initial search range is narrowed to the target accuracy; calculate and obtain the optimal weight parameters ω and β.

In order to achieve the dual purpose of contrast enhancement and brightness preservation, we first need to define an objective function J, as shown below:

和I_in,m分别是增强图像和输入图像的平均亮度，H表示熵值；亮度误差作为目标函数的惩罚项，当存在亮度误差时，熵值将作为最小值输出，如果不存在亮度误差，则图像熵值将恢复为原始值；最后，需要满足条件是亮度误差应最小，目标函数J应最大化。此时的权值为最佳权重。in

and I _in,m are the average brightness of the enhanced image and the input image, respectively, and H represents the entropy value; the brightness error is used as the penalty term of the objective function. When there is a brightness error, the entropy value will be output as the minimum value. If there is no brightness error, Then the image entropy value will be restored to the original value; finally, the conditions that need to be satisfied are that the brightness error should be minimized and the objective function J should be maximized. The weight at this time is the best weight.

When the input source image is I _in (u, v), and the average brightness value of the input image is I _{in, m} , using formula (12) and formula (13), the golden section algorithm searches for the specific implementation of the weight parameter algorithm steps Can be as follows:

Step S210, input the golden section point ρ=0.618, and iterate the initial value α ₁ =eps, α ₂ =1-eps;

Step S211, the error range is denoted as Δα=α ₁ −α ₂ , and a smaller accuracy value τ is determined, where τ→ε=10 ⁻⁴ ;

Step S212, calculating the objective functions J ₁ , J ₂ corresponding to α ₁ , α ₂ ;

Step S213, when Δα>τ and J ₁ >J ₂ are satisfied, update the interval endpoint formula, set α ₂ =α ₁ +ρ×Δα, and record J ₂ as the objective function; otherwise, update the interval endpoint formula, set α ₁ =α ₁ +(1-ρ)×Δα, and J ₁ is recorded as the objective function;

Step S214, continuously update interval endpoints α ₁ , α ₂ and objective functions J ₁ , J ₂ , repeat step S213, until Δα<τ is satisfied, end the search step;

Step S215, return and calculate the weight parameters ω and β, where ω=0.5×(α ₁ +α ₂ ), and β=1−ω.

Further optionally, after the weighted histogram equalization is performed on the V-channel image in the embodiment of the present invention, the contrast and detail information of the image can be greatly improved, but when the image is converted back to the RGB space, desaturation or color loss may also occur. Phenomenon. Through step S114, the gray value of the histogram corresponding to the S(u,v) image can be maximized and stretched, so that the distribution of the histogram corresponding to the S(u,v) image is uniform, and the enhanced S _out (u ,v) image.

In HSV space, its brightness and saturation are equations (14) and (15), respectively:

V=max{R,G,B} (14)

Where R, G and B are the normalized values of RGB, when the V channel image is enhanced, the image pixel intensity will tend to L-1, so there is V=max{R,G,B}=(L-1), The formula is modified or expressed as follows:

R(u,v)=(L-1), G(u,v)=(L-1), B(u,v)=(L-1)(16)

According to equation (15), the minimum saturation is as follows:

Among them, the higher the saturation of the image, the more color types are displayed. In order to display more color information, the saturation information needs to be expanded to the maximum range:

S _out (u,v)=max{S _in (u,v)} (18)

The image saturation stretching process is shown in Figure 5. Figure 5b in Figure 5 shows that after the saturation is stretched, the corresponding image recovers more detailed information.

Finally, in step S115, the obtained final V channel enhanced image V _out (u, v) and the final S channel enhanced image S _out (u, v), as well as the original H (u, v) can be processed as HSV and RGB The final enhanced image I _out (u, v) is obtained by inverse color space transformation, and the transformation function is shown in formula (19):

The final enhanced image I _out (u, v) is:

I _out (u,v)={R _out (u,v),G _out (u,v),B _out (u,v)} (20)

In the simulation experiments, the method proposed in the embodiment of the present invention is compared with the traditional histogram equalization method (HE), the contrast-limited adaptive equalization method (CLAHE), the average histogram equalization method (AvHeq) and the maximum coverage of the histogram. Method (MaxCover) to compare the effect, please refer to Figure 6-Figure 13 together.

FIG. 6 and FIG. 7 are a comparison graph of the enhancement result of the scene graph named "earth" and the corresponding histogram comparison reference graph. Figure 6a and Figure 7a are the source image and the corresponding histogram, respectively. Figures 6b and 7b are the enhancement results and the corresponding histogram results obtained by the traditional HE method, respectively. Figures 6c and 7c are the enhancement results and the corresponding histogram results obtained by the sampling CLAHE method, respectively. Figures 6d and 7d are the enhancement results and the corresponding histogram results obtained by the AvHeq method, respectively. Figures 6e and 7e are the enhancement results and the corresponding histogram results obtained by the MaxCover method, respectively. Figures 6f and 7f are the enhancement results and the corresponding histogram results obtained by the method of the present invention, respectively.

It can be seen from Figure 6b, Figure 6c, Figure 6d and Figure 6e that all four methods can improve the image quality. Although the overall contrast of the source image in Figures 6b and 6c is improved and the texture inside the rock is also revealed, the histograms corresponding to these two methods in Figures 7b and 7c show that the shape of the histogram of the source image does not is effectively reserved. Part of the detailed color information of the image shown in Fig. 6d is not recovered efficiently. Figures 6e and 7e show that the image details are seriously lost and the overall effect is blurred, and the corresponding histogram results are not evenly distributed to the entire interval. Figures 6f and 7f are the results obtained by using the method of the present invention, and the results show that the texture detail information of the source image is displayed more clearly. Also, the histogram results are better. It can be seen that the method provided by the embodiment of the present invention can better improve the image contrast and effectively restore the color information of the source image.

FIG. 8 and FIG. 9 are a comparison diagram of an enhancement result of the present invention for a scene graph named "road" and a corresponding histogram comparison reference diagram. 8a and 9a are the source image and the histogram corresponding to the source image, respectively. Figure 8b and Figure 9b are the enhancement results obtained by the traditional HE method and the corresponding histogram results, respectively. Figures 8c and 9c are the enhancement results and the corresponding histogram results obtained by the sampling CLAHE method, respectively. Figure 8d and Figure 9d are the enhancement results and the corresponding histogram results obtained by the AvHeq method, respectively. Figures 8e and 9e are the enhancement results and the corresponding histogram results obtained by using the MaxCover method, respectively. Fig. 8f and Fig. 9f are respectively the enhancement result obtained by the method of the present invention and the corresponding histogram result.

As shown in Figures 8b and 8c, the HE method and the CLAHE method have obvious dehazing effect on the source image, but the color information is not restored. Figure 8d, Figure 8e, Figure 9d, and Figure 9e show that the AvgHe method and the MaxCover method can effectively preserve the shape of the source image histogram, but the dehazing effect and color information recovery effect are not satisfactory. The result of the embodiment of the present invention is shown in Fig. 8f, the subjective visual effect is the best, especially the color information of the houses, cars on the left and right sides, and the trees in the distance are restored obviously, and the details are also effectively enhanced.

FIG. 10 and FIG. 11 are the comparison diagrams of the enhancement results of the present invention for the scene graph named "126007" and the corresponding histogram comparison reference diagrams. Figure 10a and Figure 11a are the source image and the corresponding histogram, respectively. Figure 10b and Figure 11b are the enhancement results and the corresponding histogram results obtained by the traditional HE method, respectively. Figure 10c and Figure 11c are the enhancement results and the corresponding histogram results obtained by the sampling CLAHE method, respectively. Figure 10d and Figure 11d are the enhancement results and the corresponding histogram results obtained by the AvHeq method, respectively. Figures 10e and 11e are the enhancement results and the corresponding histogram results obtained by using the MaxCover method, respectively. FIG. 10f and FIG. 11f are the enhancement results and the corresponding histogram results obtained by the method of the present invention, respectively.

Figures 10b and 10c, the HE method and the CLAHE method are used to distort part of the sky color of the source image; Figures 10d and 10e show that the AvgHe method and the MaxCover method do not obtain satisfactory brightness preservation effect for the source image, and the overall contrast of the image is still However, the result of the embodiment of the present invention is shown in Fig. 10f, no color distortion occurs in the source image, and the contrast is greatly improved while maintaining the brightness.

FIG. 12 and FIG. 13 are a comparison diagram of the enhancement result of the present invention for a scene graph named "5096" and a corresponding histogram comparison reference diagram. Figure 12a and Figure 13a are the source image and the corresponding histogram, respectively. Figure 12b and Figure 13b are the enhancement results and the corresponding histogram results obtained by the traditional HE method, respectively. Figure 12c and Figure 13c are the enhancement results and the corresponding histogram results obtained by the sampling CLAHE method, respectively. Figure 12d and Figure 13d are the enhancement results and the corresponding histogram results obtained by the AvHeq method, respectively. Figure 12e and Figure 13e are the enhancement results and the corresponding histogram results obtained by using the MaxCover method, respectively. Fig. 12f and Fig. 13f are respectively the enhancement result and the corresponding histogram result obtained by the method of the present invention.

As shown in Figure 12b and Figure 12c, the HE method and the CLAHE method are used to enhance the ground part of the source image, and the sky color is distorted. Figures 12d and 12e show that the AvgHe method and the MaxCover method have a good effect on the restoration of source image detail information. However, the result of the embodiment of the present invention is shown in Fig. 12f, the color restoration effect of the wall, brick and sky color information of the source image is satisfactory, and the overall brightness information is also well improved.

The evaluation of image enhancement results is divided into subjective evaluation and objective evaluation. The subjective evaluation directly observes the enhancement effect of the brightness information, contrast information and color information of the experimental result image through the visual system. Objective evaluation is to use image statistical parameters to make judgments. The test results of the present invention can adopt entropy value, image clarity (Tenengrad gradient) and average gradient index as objective parity data content.

The image entropy value can characterize the amount of image information. The larger the entropy value, the richer the retained detail information. Formula (13) can be used as the test basis for entropy value.

Image sharpness reflects the overall visual effect of the image, and higher sharpness test results indicate better subjective visual quality of the image. It can be expressed by formula (21):

where T is the threshold, and Δm _x (u,v) and Δnx( _u ,v) are the differences between the pixels in the horizontal and vertical directions of the pixel (u,v), respectively.

The average gradient describes the richness of image details, and a higher average gradient test result indicates the richer the image details. It can be expressed by formula (23):

和

表示行和列中的可变像素数量。四组增强实验的熵值、清晰度和平均梯度客观数据结果如表格1所示：where f(u,v) is the gray level of the pixel,

and

Represents a variable number of pixels in rows and columns. The objective data results of entropy, sharpness and average gradient of the four groups of enhancement experiments are shown in Table 1:

Table 1 Objective evaluation index data of different algorithms

It can be obtained from the data in Table 1 that in the comparison of the four groups of image enhancement result parameters, the results of the sharpness data and the average gradient data obtained by the method of the embodiment of the present invention are both the largest, indicating that the enhancement effect is the best. The entropy data results of Figure 6, Figure 8 and Figure 12 also achieve the largest data results compared to other enhancement algorithms. Although the entropy value test result of the AvHeq algorithm in FIG. 10 is 7.7242, which is higher than that of the method proposed in the present invention, it has been qualitatively shown that the method proposed in the embodiment of the present invention has higher performance than the AvHeq algorithm in terms of visual perception effect, The comprehensive effect is better, and better results can be obtained compared with other schemes in terms of image brightness preservation and color information recovery.

In the embodiment of the present invention, the R, G, and B channel color amplitudes of the source image are first stretched, and then the V channel image and the S channel image are adjusted based on the HSV space, so as to equalize the histogram of the V channel image and maximize the gray value of the S channel image. stretch, and then inversely transform the adjusted image into RGB space to obtain an enhanced image. The method provided by the embodiment of the present invention overcomes the “over-enhancement” phenomenon caused by the traditional histogram equalization algorithm, and the calculation process is concise and complicated. It can effectively improve the contrast and brightness information of the image, and obtain an enhanced image with high contrast and high color brightness.

Referring to FIG. 14 , an embodiment of the present invention further provides an image enhancement apparatus, which can be used to implement the methods shown in FIG. 1 and FIG. 2 . Specifically, the image enhancement apparatus provided by the embodiment of the present invention includes:

Obtaining unit 11, for obtaining the source image I _in (u, v), I _in (u, v)={R _in (u, v), G _in (u, v), B _in (u, v)} ,; the specific implementation can refer to the aforementioned step S110;

The RGB stretching unit 12 is used to stretch the color amplitudes of the R, G and B channels of I _in (u, v) to obtain the image I _stre (u, v) after the color amplitude is stretched, I _stre (u, v) ={R _stre (u, v), G _stre (u, v), B _stre (u, v)}; for the specific implementation, refer to the aforementioned step S111;

The conversion unit 13 is used to convert the color space of I _stre (u, v), so that I _stre (u, v) is converted from the RGB space to the HSV space to obtain the hue component H(u, v), the luminance component V(u, v) and the saturation component S(u, v); the specific implementation can refer to the aforementioned step S112;

The calculation unit 14 is used for weighted calculation V(u, v) to obtain a new V channel image V _out (u, v); for a specific implementation, refer to the aforementioned step S113;

The saturation component stretching unit 15 is used to stretch S(u, v) to obtain the stretched S channel image S _out (u, v); for the specific implementation, refer to the aforementioned step S114;

The inverse transform unit 16 is used for inversely transforming H(u,v), _Vout (u,v) and _Sout (u,v) to RGB space to obtain an enhanced image _Iout (u,v), I _out (u, v)={R _out (u, v), G _out (u, v), B _out (u, v)}; the specific implementation can refer to the aforementioned step S115.

Further optionally, as shown in FIG. 15 , the RGB stretching unit 12 provided by the embodiment of the present invention may include an R calculation unit 121 , an R selection unit 122 , and an R stretching unit 123 , which are used to specifically stretch the color of the R channel. Amplitude:

R calculation unit 121, for calculating the R channel color amplitude value range of I _in (u, v);

R selection unit 122, for selecting the highest value in the calculated R channel color amplitude value range;

The R stretching unit 123 is used to stretch the color amplitude of the R channel to the highest value in the range of the color amplitude of the R channel to obtain R _stre (u, v);

Further optionally, as shown in FIG. 16 , the RGB stretching unit 12 provided in the embodiment of the present invention may include a G calculation unit 124, a G selection unit 125, and a G stretching unit 126, which are used to specifically implement the stretching of the G channel color. Amplitude:

The G calculation unit 124 is used to calculate the G channel color amplitude value range of I _in (u, v);

G selection unit 125 is used to select the highest value in the calculated G channel color amplitude value range;

The G stretching unit 126 is used to stretch the color amplitude of the G channel to the highest value in the range of the color amplitude of the G channel to obtain G _stre (u, v);

Further optionally, as shown in FIG. 17 , the RGB stretching unit 12 provided by the embodiment of the present invention may include a B calculation unit 127 , a B selection unit 128 , and a B stretching unit 129 for specifically implementing the stretching of the B channel color. Amplitude:

B calculation unit 127, for calculating the B channel color amplitude value range of I _in (u, v);

B selection unit 128 is used to select the highest value in the calculated B channel color amplitude value range;

The B stretching unit 129 is used to stretch the color amplitude of the B channel to the highest value in the value range of the color amplitude of the B channel to obtain B _stre (u, v).

Further optionally, as shown in FIG. 18 , the calculation unit 14 provided in this embodiment of the present invention may include a weighted calculation unit 141, a normalization unit 142, and a mapping unit 143 for implementing weighted equalization calculation:

The weighted calculation unit 141 is configured to perform weighted calculation on the gray value of V(u, v) corresponding to the histogram h(i) to obtain the weighted histogram _hW (i), _hW (i)=β×h (i)+ω×max{(h(i))}, 0<ω<1, ω+β=1, max{h(i)} is the maximum gray value;

The normalization unit 142 is used to normalize h _W (i) to obtain h _N (i), wherein,

The mapping unit 143 is configured to map the non-null intensity levels in h _N (i), so that the histogram distribution of h _N (i) is uniform, and the weighted V _out (u, v) is obtained.

Further optionally, as shown in FIG. 19 , the weighting calculation unit 141 provided in this embodiment of the present invention may further include a selection unit 1411 , an iteration and reduction unit 1412 , and an optimal weight parameter calculation unit 1413 , for implementing the optimal Selection of weight parameters ω and β:

The selection unit 1411 is used to select the initial search range; it is also used to select the luminance error as the iterative objective function;

The iteration and narrowing unit 1412 is used to iterate through the objective function to narrow the initial search range; and is also used to repeat the iterative operation through the narrowed initial search range until the initial search range is narrowed to the target precision;

The optimal weight parameter calculation unit 1413 is configured to calculate and obtain the optimal weight parameters ω and β.

The saturation component stretching unit 15 in the embodiment of the present invention can be specifically used to maximize the gray value of the intensity level of the histogram corresponding to the S(u,v) image, so that the histogram corresponding to the S(u,v) image can be stretched to the maximum extent. The image distribution is uniform, and the enhanced S _out (u, v) image is obtained. For a specific implementation manner, reference may be made to the foregoing steps, which will not be repeated here.

The image enhancement device of the embodiment of the present invention firstly stretches the color amplitude of R, G and B channels of the source image, and then adjusts the V channel image and the S channel image based on the HSV space, so that the histogram of the V channel image is equalized and the S channel image is grayed out. The degree value is maximized and stretched, and then the adjusted image is inversely transformed into RGB space to obtain an enhanced image. The image enhancement device provided by the embodiment of the present invention overcomes the "over-enhancement" phenomenon caused by the traditional histogram equalization algorithm. , the calculation process is simple, the complexity is low, and the contrast and brightness information of the image can be effectively improved, and an enhanced image with high contrast and high color brightness is obtained.

The modules or units in the embodiments of the present invention may be implemented by a general-purpose integrated circuit, such as a CPU (Central Processing Unit, central processing unit), or an ASIC (Application Specific Integrated Circuit, an application-specific integrated circuit).

Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be implemented by instructing relevant hardware through a computer program, and the program can be stored in a computer-readable storage medium. During execution, the processes of the embodiments of the above-mentioned methods may be included. The storage medium may be a magnetic disk, an optical disk, a read-only memory (Read-Only Memory, ROM), or a random access memory (Random Access Memory, RAM) or the like.

The steps in the method of the embodiment of the present invention may be adjusted, combined and deleted in sequence according to actual needs.

The modules or units in the apparatus of the embodiment of the present invention may be combined, divided, and deleted according to actual needs.

What is disclosed above is only the preferred embodiment of the present invention, of course, it cannot limit the scope of the right of the present invention. Those of ordinary skill in the art can understand that all or part of the process of realizing the above-mentioned embodiment can be made according to the claims of the present invention. The equivalent changes of the invention still belong to the scope covered by the invention.

Claims (10)

1. an image enhancement method, is characterized in that, described method comprises: Obtain the source image I _in (u, v), the I _in (u, v)={R _in (u, v), G _in (u, v), B _in (u, v)}; Stretch the R, G, and B channel color amplitudes of the I _in (u, v) to obtain an image I _stre (u, v) after the color amplitude is stretched, the I _stre (u, v)={R _stre (u,v),G _stre (u,v),B _stre (u,v)}; Convert the color space of the I _stre (u, v) to convert the I _stre (u, v) from the RGB space to the HSV space to obtain the hue component H(u, v) and the luminance component V(u, v) and the saturation component S(u,v); Weighted calculation of the V(u, v) to obtain a new V channel image V _out (u, v); Stretch the S(u, v) to obtain the stretched S channel image S _out (u, v); Inversely transform the H(u,v), the _Vout (u,v) and the _Sout (u,v) to the RGB space to obtain an image enhanced image _Iout (u,v), the _Iout (u,v)={ _Rout (u,v), _Gout (u,v), _Bout (u,v)}. 2. The method according to claim 1, wherein the R, G and B channel color amplitudes of the stretched I _in (u, v) specifically include: Calculate the R channel color amplitude value range of described I _in (u, v); Select the highest value in the calculated R channel color amplitude value range; Stretch the R channel color amplitude to the highest value in the R channel color amplitude value range to obtain the R _stre (u, v); Calculate the range of values of the G channel color amplitude of the I _in (u, v); Select the highest value in the calculated G channel color amplitude value range; Stretch the G channel color amplitude to the highest value in the G channel color amplitude value range to obtain the G _stre (u, v); Calculate the B channel color amplitude value range of the I _in (u, v); Select the highest value in the calculated B channel color amplitude value range; The B channel color amplitude is stretched to the highest value in the value range of the B channel color amplitude to obtain the B _stre (u, v). 3. The method of claim 1, wherein the weighted calculation V(u, v) obtains a new V channel image V _out (u, v), specifically comprising: Perform weighted calculation on the gray value of the histogram h(i) corresponding to the V(u,v) to obtain the weighted histogram _hW (i), the _hW (i)=β×h(i) +ω×max{(h(i))}, 0<ω<1, ω+β=1, the max{h(i)} is the maximum gray value; Normalize the _hW (i) to obtain _hN (i), where,

The non-empty intensity levels in the h _N (i) are mapped to make the histogram distribution of the h _N (i) uniform, and the V _out (u, v) after the weighting calculation is obtained. 4. method as claimed in claim 3, is characterized in that, described by described V (u, v) the gray value of corresponding histogram h (i) is weighted and calculated, obtains weighted histogram h _W ( i) before, also include: Select the initial search scope; Select the luminance error as the iterative objective function; Iterating through the objective function to narrow the initial search range; Repeating the iterative operation by reducing the initial search range until the initial search range is narrowed to a target precision; The ω and the β are calculated to obtain the optimal weight parameters. 5. The method according to claim 1, wherein the stretching of the S(u, v) to obtain the stretched S channel image S _out (u, v) specifically includes: The intensity level of the histogram corresponding to the S(u,v) image is stretched to maximize the gray value, so that the distribution of the histogram corresponding to the S(u,v) image is uniform, and the enhanced S(u,v) image is obtained. _out (u,v) image. 6. An image enhancement device, characterized in that the device comprises: an acquisition unit for acquiring a source image I _in (u, v), the I _in (u, v)={R _in (u, v), G _in (u, v), B _in (u, v) }; The RGB stretching unit is used to stretch the color amplitudes of the R, G and B channels of the I _in (u, v) to obtain an image I _stre (u, v) after the color amplitude is stretched, and the I _stre (u, v) ,v)={R _stre (u,v),G _stre (u,v),B _stre (u,v)}; A conversion unit, configured to convert the color space of the I _stre (u, v), so that the I _stre (u, v) is converted from the RGB space to the HSV space to obtain the hue component H(u, v), the luminance component V (u,v) and the saturation component S(u,v); a calculation unit, used for weighted calculation of the V(u, v), to obtain a new V channel image V _out (u, v); a saturation component stretching unit, used to stretch the S(u, v) to obtain the stretched S channel image S _out (u, v); an inverse transform unit for inversely transforming the H(u,v), the _Vout (u,v) and the _Sout (u,v) to the RGB space, to obtain an image enhanced image _Iout (u , v), the I _out (u, v)={R _out (u, v), G _out (u, v), B _out (u, v)}. 7. The device of claim 6, wherein the RGB stretching unit comprises: R calculation unit, for calculating the R channel color amplitude value range of the I _in (u, v); R selection unit, for selecting the highest value in the calculated R channel color amplitude value range; The R stretching unit is used to stretch the color amplitude of the R channel to the highest value in the range of the color amplitude of the R channel to obtain the R _stre (u, v); G calculation unit, for calculating the G channel color amplitude value range of the I _in (u, v); G selection unit, for selecting the highest value in the calculated G channel color amplitude value range; G stretching unit, used for stretching the color amplitude of the G channel to the highest value in the value range of the color amplitude of the G channel, to obtain the G _stre (u, v); B calculation unit, for calculating the B channel color amplitude value range of the I _in (u, v); B selection unit, for selecting the highest value in the calculated B channel color amplitude value range; The B stretching unit is used to stretch the color amplitude of the B channel to the highest value in the value range of the color amplitude of the B channel to obtain the B _stre (u, v). 8. The apparatus of claim 6, wherein the computing unit comprises: A weighted calculation unit, configured to perform weighted calculation on the gray value of the histogram h(i) corresponding to the V(u, v) to obtain the weighted histogram _hW (i), where the _hW (i)= β×h(i)+ω×max{(h(i))}, 0<ω<1, ω+β=1, the max{h(i)} is the maximum gray value; A normalization unit for normalizing the _hW (i) to obtain _hN (i), where,

A mapping unit, configured to map the non-empty intensity levels in the h _N (i), so that the histogram distribution of the h _N (i) is uniform, and the weighted V _out (u, v) is obtained. 9. The apparatus according to claim 8, wherein the weight calculation unit comprises: a selection unit for selecting an initial search range; also for selecting a luminance error as an iterative objective function; an iterative and narrowing unit for iterating through the objective function to narrow the initial search range; also using repeating the iterative operation through the narrowed initial search range until the initial search range is narrowed to the target precision; The optimal weight parameter calculation unit is configured to calculate the ω and the β to obtain the optimal weight parameter. 10 . The apparatus according to claim 6 , wherein the saturation component stretching unit is specifically configured to maximize the stretching of the gray value of the histogram corresponding to the S(u,v) image, 10 . The histogram corresponding to the S(u,v) image is uniformly distributed to obtain the enhanced S _out (u,v) image.

Images