KR101071567B1 - Contrast Enhancement Using a Density based Sub-histogram Equalization Technique - Google Patents

Abstract

The present invention relates to a method for enhancing contrast of an image by dividing the histogram according to density characteristics and smoothing the histogram for each region. The histogram is divided into regions of high density, taking into account the histogram density and the amount of histogram accumulation. When regions with similar characteristics are found, histogram smoothing is performed using the brightness information included in each region. Finally, by combining the smoothed result images for each area, the contrast quality is improved as much as possible without excessive smoothing distortion, thereby improving the quality of the input image.

 Density-based, histogram, smoothing, cumulative amount, contrast

Description

Contrast Enhancement Using a Density based Sub-histogram Equalization Technique}

Communication

Contrast enhancement methods are divided into global contrast enhancement methods that refer to the image of the entire image and local contrast enhancement methods that refer to the image of the block image made by dividing the entire image into several areas. . One method of improving global contrast is the Global Histogram Equalization. Global histogram smoothing method is to calculate the brightness distribution histogram (PDF) of the image with reference to the whole image, and redistribute the contrast so that the brightness distribution histogram has a uniform distribution by generating and normalizing the cumulative distribution function (CDF). . However, since global histogram smoothing uses histogram information of the entire image without considering sub-space information, it is difficult to calculate local contrast values. In addition, since the normalization is performed uniformly over all the brightness areas without considering the density of the histogram, when the brightness value of a specific area is very dense, there are many distortions that are excessively smoothed during the normalization process, and even the average brightness value It changes badly.

In order to solve this problem, Chen proposed a method of recursive mean-separate histogram equalization (RMSHE) that repeatedly divides and smooths histograms based on average brightness to preserve the average brightness of the original image. However, this method has a disadvantage in that the number of histograms to be divided according to the image is not clear, and it is not flexible because the number of repetitions is also an exponential power interval of 2. On the other hand, in order to overcome the problems of the global contrast enhancement method, contrast enhancement methods have been proposed, in which an entire image is divided into several block images and smoothed with only the image of each block. Adaptive Histogram Equalization (AHE) is the most basic method that divides the image into several blocks without overlapping and performs histogram smoothing on each block independently and combines the results by bilinear interpolation. In this case, in order to solve the discontinuity problem that occurs at the part where the blocks are connected, the neighboring blocks are divided so that they overlap.

However, the degree of blockage in which discontinuities appear depending on the selection of overlapping block sizes varies, and there is a problem in that the size of blocks must be determined manually for each image. In general, a common problem with local contrast enhancement methods is blocking. This blocking phenomenon refers to a problem in that the contrast between blocks of the image is discontinuously displayed because only the image information of the block image is used without considering the brightness distribution of the entire image. In order to minimize this phenomenon, blocks are overlapped or interpolation is used, resulting in large computational complexity or damage to features in the resulting block image.

In order to solve the above problems of the contrast enhancement method, the present invention distinguishes high density areas causing excessive smoothing by analyzing histogram distribution of the input image, and performs histogram smoothing for each range of divided brightness values. By doing so, we want to improve the global contrast of the input image obtained from the camera.

According to the present invention, a method for improving contrast of an image through smoothing of histogram regions according to the present invention includes: (a) dividing a high density region that causes distortion through histogram analysis; (b) performing histogram smoothing on each of the divided histogram regions, and combining the resulting images to obtain an image having improved contrast without distortion.

Step (a) may include generating a histogram on the input image and calculating a density of the histogram by referring to the histogram of the ambient brightness value; Arranging the histogram density in the order of high brightness value; Expanding the dense histogram area around a brightness value indicating a similar high density value; A step of finding a histogram region having a high density, which includes finding a boundary value by approximating a region measured at a high density with a Gaussian distribution.

Step (b) comprises the steps of: performing histogram smoothing for each area divided in step (a); By adding a histogram smoothing image for each region smoothed in different ranges, it is a method of improving contrast globally without distortion of the image.

The present invention relates to a method for improving contrast of an image by segmenting the histogram according to density characteristics and smoothing the divided histogram by region. The biggest problem of the global contrast enhancement method is to perform smoothing without considering the histogram distribution of the image and to damage the image due to excessive smoothing of the dense area. In order to solve this problem, the histogram area having a high density is divided in consideration of the histogram density and the cumulative amount of the histogram. When regions with similar characteristics are found, histogram smoothing is performed using the brightness information included in each region. Lastly, by smoothing the resultant images for each area, the image quality of the input image is improved by increasing the contrast as much as possible without excessive smoothing distortion.

In accordance with another aspect of the present invention, there is provided a method for improving contrast through a density-based segmented histogram smoothing method comprising: (a) dividing a high-density region causing distortion through histogram analysis according to brightness values of an image; (b) performing histogram smoothing on each of the divided histogram regions and combining the resulting images to obtain an image having improved contrast without distortion. Figure 1 shows the configuration of the contrast enhancement method through the density-based segmented histogram smoothing of the invention.

The density value Dm given in Equation 1 is calculated by using the sum of the distribution amounts of neighboring brightness values in each brightness value as the first process to distinguish the areas with high density through the histogram analysis in step (a). do. In Equation 1, w denotes a range of neighbors to be referred to, m denotes a center brightness value to be obtained, and f ( xi ) denotes a frequency at the i- th brightness value. 2 shows a graph representing the density value obtained at each brightness value.

The next step is to sort the calculated density values in order of high density in order to construct the regions in the order of density. 3 shows the results of sorting in the order of high density through numbering on the density value graph.

4 is a flowchart illustrating a method of widening an area around a brightness value having a high density. When the density values are aligned, the average density value is compared with the density at the neighboring brightness value at the brightness value point indicating the high density to form the area around the area showing the high density value (13). At this time, if the difference between the density values of the average density value and the neighboring brightness value is less than or equal to the threshold value, the neighboring brightness value area is determined to be the same area, and after updating the average density value when these neighboring brightness values are included, (12), the comparison with the next neighboring brightness value is repeated (11). However, if there is a large difference in density values, the neighboring region is determined to be another region and the expansion of the region is stopped (14). FIG. 5 shows that regions are formed around a brightness value having a high density in the density value graph obtained in FIG. 2.

As the final step of dividing the region, the distribution of the divided region is approximated by Gaussian distribution to set the point where each distribution overlaps or becomes zero as the boundary of the region. Equation 2 represents a Gaussian function for approximation, where μ _n is a center point of each region divided in the previous step, and σ is obtained through a process of obtaining variance as shown in Equation 3. FIG. 6 shows that the boundary of the region is determined by approximating the distribution of the regions formed in FIG. 5 to a Gaussian distribution.

The method of improving contrast by smoothing the divided histogram areas of step (b) uses the histogram area information obtained from the result of step (a) to obtain an image with improved contrast without distortion in each area. Can be. FIG. 7 shows a result image of histogram smoothing performed for each region when the image as shown in (a) is divided into three regions as shown in (b) through (d), (e), and (f). .

When the image of each region with improved contrast is obtained without distortion due to excessive smoothing, the image with improved contrast can be finally obtained by combining the images. FIG. 8 illustrates the result of improved contrast on a global basis by adding a smoothed image for each region in three regions of FIG. 7.

1 is a block diagram of a method for improving contrast through the divided histogram smoothing of the density-based according to the present invention.

2 is a density graph of brightness values of an input image of the present invention.

3 is a graph showing brightness values arranged in density order of an input image of the present invention.

4 is a block diagram of a method of forming an area at a brightness value having a high density of an input image of the present invention.

5 is a graph showing the formation of an area at a high brightness value of the input image of the present invention.

6 is a graph showing an approximation to a Gaussian distribution of a region formed based on the density value of the present invention.

7 is an image showing a result of performing histogram smoothing for each region with reference to the boundary of the found region of the present invention.

8 is a final result image obtained by improving the contrast by combining the histogram smoothed images for each region of the present invention.

Claims (6)

delete delete delete In the contrast enhancement method based on density-based segmented histogram smoothing, Step 1-1 (S1-1) of generating a histogram on the input image acquired by the camera, calculating a density of the histogram by referring to the histogram of the ambient brightness value, and arranging the histogram density in the order of high brightness values. Step 2 (S1-2) and a high density value, wherein the high density value indicates a density value of the inflection point at which the density value increases and decreases. The cause of distortion through histogram analysis includes steps 1-3, and steps 1-4 (S1-4) of finding a boundary value by approximating a distribution of the extended histogram region with a Gaussian distribution. A first step (S1) of classifying regions of high density; And a second step (S2) of performing histogram smoothing for each of the divided areas in the first step (S1), and combining the resulting images to obtain an image having improved contrast without distortion. The method of claim 4, wherein The density value Dm of the first step S1-1 is

Wherein w denotes the range of neighbors to refer to, m denotes the center brightness value to be obtained, and f (xi) denotes the frequency at the i-th brightness value. The method of claim 4, wherein The first step (S1-3) is When the density values are aligned, steps 1-3a of comparing an average density value at a brightness point indicating the high density value with a density at a neighboring brightness value to form an area around the area indicating the high density value (S1-3a); If the difference in density between the average density value and the neighboring brightness value is less than or equal to the threshold value, the neighboring brightness value area is determined as the same area, and the first density for updating the average density value when these neighboring brightness values are included is determined. Step -3b (S1-3b); (1-3c) step S1-3c of repeating the comparison with the neighboring brightness value after updating the average density value; If the difference in density value between the average density value and the neighboring brightness value is greater than or equal to the threshold value, the neighboring area is determined to be another area and includes steps 1-3d (S1-3d) of stopping expansion of the area. Features a contrast enhancement method.

Images