TWI595449B - Image enhance method - Google Patents

Description

Image enhancement method

The invention is an image enhancement method, in particular to an image enhancement method capable of retaining an original hue of an image while increasing contrast.

Since digital imaging has become the mainstream of imaging media, the development of image processing has emerged. Image processing is a general term for digitally or analogous media to define an infinite number of pixels in a digital manner and to make various modulations for the color parameters of each pixel.

Since the image processing is performed on the target pixel by the same rule, the processed image is often inevitably caused to have defects such as excessive contrast, reduced chroma or over saturation, high color temperature, cold or noise enhancement. In a nutshell, the key to image processing quality is not only how to highlight a specific image, but also how to achieve the above goal without excessive distortion. This is also a difficult point for image processing.

One of the conventional image enhancement methods is the histogram equalization method, which is used to enhance the global contrast of an image. However, the method is to accumulate the grayscale values of all the pixels without any difference and equalize them, so it is quite easy to enhance the noise in the image. In addition, there may be local areas in the original image that are subject to noise. Since the brightness is deviated by the influence, after the histogram equalization method, there is a possibility that the regions having the same color actually form different colors.

In addition, the traditional histogram equalization method does not consider the hue retention, so after the contrast enhancement in this way, although the contour of the image becomes more obvious due to the global contrast enhancement, only the grayscale value difference of each pixel is applied. Not too big, if the grayscale value difference between pixels is large, the hue of the processed image will be seriously offset and distorted.

In view of the defects caused by the prior art in image enhancement, the present invention proposes an image enhancement method capable of retaining the hue of the original image, and it is desirable to solve the problem of causing serious deviation of the hue after the prior art enhanced image.

According to an embodiment of the present invention, an image enhancement method includes the steps of: reading an original image and converting the original image into a first grayscale image. A Gabor filter is used to enhance the first grayscale image to form a second grayscale image. Combining the first grayscale image and the second grayscale image to form a composite grayscale image. Obtain one of the image parameters of the composite grayscale image. The histogram equalization step is performed on the composite gray scale image by using the image parameters. A correction value is obtained by comparing the intensity values of the composite grayscale image and the first grayscale image after the histogram equalization step. And the original image is enhanced by the correction value to obtain a first-class image.

As can be seen from the above, in the present embodiment, instead of directly using the histogram equalization to enhance the original image, the correction value is generated by using the characteristic relationship between the composite grayscale image and the grayscale image of the original image. Thus, the present embodiment only relies on The correction value processes the original image to preserve the hue of the original original image.

In the foregoing embodiment, the correction value may be a ratio or a difference between the intensity values of the composite grayscale image and the first grayscale image; and in the case where the correction value is the two different calculation modes, the hue values of the equalized image are respectively It can be the product of the hue value of the original image and the correction value and the sum of the additions. The aforementioned image parameters may include a probability density function and a cumulative distribution function, but are not limited thereto. The foregoing Gabor filter can enhance the first grayscale image by a calculation program to form a second grayscale image, and the calculation program can be: Where u=x cos θ+y sin θ, v=-x sin θ+y cos θ, θ is the direction of the Gabor filter, and σ _u and σ _v are the standard deviations of the Gaussian function on the u and v axes, respectively , ω is the modulation frequency.

Therefore, in the embodiment, the correction values of the intensity values of the generated composite gray scale image and the first gray scale image are differently strengthened, and the flexibility of the enhancement result of the equalized image is retained.

100‧‧‧Image enhancement method

110~170‧‧‧Steps

1 is a flow chart showing the steps of an image enhancement method according to an embodiment of the present invention; FIG. 2A is a color image view of the original image; and FIG. 2B is a first gray-scale image view of the original image; 2C is a histogram equalization color image of the first grayscale image of FIG. 2B; FIG. 3A is a second grayscale of the first grayscale image of FIG. 2B after Gabor filtering. Image; Figure 3B shows the composite grayscale image of Figure 2B and Figure 3A; Figure 4A shows the color image of the first equalized image; Figure 4B shows the second image. Color image of the image; Figure 5A shows the color image of the retinal image; Figure 5B shows the color image of the first equalized image of Figure 5A; and Figure 5C shows the 5A The color image of the second equalized image of the figure.

Please refer to FIG. 1 , which is a flow chart showing the steps of the image enhancement method 100 according to an embodiment of the present invention. Image enhancement method 100 includes steps 110 through 170. Step 110 is to read an original image and convert the original image into a first grayscale image. Step 120 is to use a Gabor filter to enhance the first grayscale image to form a second grayscale image. Step 130 is to combine the first grayscale image and the second grayscale image to form a composite grayscale image. Step 140 is to obtain one of the image parameters of the composite grayscale image. Step 150 is a histogram equalization step of the composite grayscale image by using the image parameters. Step 160 is to obtain a correction value by comparing the intensity values of the composite grayscale image and the first grayscale image after the histogram equalization step. Step 170 is to enhance the original image by using the correction value to obtain the first-class image. image.

Please refer to the 2A, 2B, and 2C drawings together. Figure 2A shows a color image of the original image. Figure 2B shows the first grayscale image of the original image. 2C is a histogram equalized color image of the first grayscale image of FIG. 2B. The above step 110 is a grayscale common in conventional image processing, so the implementation details of FIG. 2B are not described here. As shown in FIG. 2C, in the prior art, the original image of FIG. 2A is directly subjected to histogram equalization processing according to FIG. 2B of the gray-scaled image, and the processed image appears to have a clear outline of the line due to the contrast increase, but the hue is The difference from Figure 2A is too large, that is, the hue is deviated.

Step 120 is to perform image enhancement on the first grayscale image by using a Gabor filter. The calculation procedure of the Gabor filter is: The above x and y are coordinates defined by the respective pixels in the original image. Where u=x cos θ + y sin θ , v =- x sin θ + y cos θ , θ is the direction of the Gabor filter, and σ _u and σ _v are the standard deviations of the Gaussian function on the u and v axes, respectively , ω is the modulation frequency.

FIG. 3A is a second gray-scale image after Gabor filtering of the first gray-scale image of FIG. 2B. Referring to FIG. 3A, the second grayscale image is obtained by volume-integrating the first grayscale image of FIG. 2B with the Gabor filter, and the equation is: Where G(x, y) is the original image, (x, y) is a filtered filter image, and h(n, m) is a filter template of M×N size. Filtered image (x, y) is as shown in Figure 3A.

In detail, the original image is an original image to be processed. In the general histogram equalization, the number of occurrences of each grayscale value in the first grayscale image is directly counted, and a histogram is drawn and directly performed. Equalization, the benefits of these processing methods are fast and reversible, the disadvantage is that the characteristics of the local pixels are not considered, and the original image may cause serious hue deviation after equalization.

In this embodiment, the first grayscale image of the original original image is filtered by the Gabor filter to generate the second grayscale image. The advantage of the Gabor filter is that the detail portion of each part of the image can be better extracted, and the direct improvement is directly performed. The problem of hue deviation caused by graph equalization.

The composite grayscale image combines the aforementioned first grayscale image and the second grayscale image, so that the details of the first grayscale image can be made clearer. Thereafter, image parameters of the composite grayscale image, such as a probability density function (PDF) and a cumulative distribution function (CDF), are calculated for subsequent histogram equalization steps. After the histogram equalization step, the present embodiment compares the intensity values of the composite grayscale image and the first grayscale image to obtain a correction value, such as a ratio or a difference, but not limited thereto. The correction value is then used to establish an operational relationship between the equalized image and the original original image, thereby enhancing the original image.

Figure 3B shows the composite ash merged between Figure 2B and Figure 3A. Order image. In the present embodiment, FIG. 3A is a detail portion for indicating the first gray scale image map, and step 130 is a combination of the second B map and the third A map to generate a composite gray scale image map of FIG. 3B. Compared with the original grayscale image of the original 2B image, it can be seen that the detail of the 2B image is relatively blurred, and the combined composite grayscale image has a sharper line, which can clearly show the details of the image.

The image parameters mentioned in the above step 140 refer to parameters applied to the back-end histogram equalization method, such as the probability density function (PDF), which can be expressed as follows: Where L is the gray level of the image, n _k is the number of occurrences of the grayscale value k in the image, and P(k) is the probability of the grayscale value k. And the cumulative distribution function (CDF), which can be expressed as follows: C is a cumulative distribution function of gray scale values. The two image parameters introduced here are common in the histogram equalization method, so the details of the equalization process of step 150 are not described.

Please continue to refer to Figure 4A and Figure 4B. Figure 4A shows a color image of the first equalized image. Figure 4B shows a color image of the second equalized image. The manner in which the two correction values are generated in step 160 will be described below.

Assuming that the color original image is O _{(i, j)} , the first grayscale image intensity value is G _{( i , j )} , where (i, j) represents the pixel coordinates. The intensity value after the histogram equalization in step 150 is G' _{( i , j )} , and the equalized image of the final retained hue is O' _{( i , j )} . The following two correction values of the example step 160, and the corresponding equalization image generation method of the step 170, the first one is: O' _{k ( i , j )} = O _{k ( i , j )} ×R; in this way, the equalized image and the original image are regarded as a multiplicative product relationship. In detail, the first way is to combine the gray scale image. The intensity value relative to the first grayscale image is regarded as an image enhancement ratio R (each pixel has a different ratio after calculation), and after the correction value is obtained, it is multiplied with the original image, and The corrected image O' _{( i , j )} whose hue is corrected is obtained, and the result is as shown in Fig. 4A.

The second way is: D = G' _{( i , j )} - G _{( i , j )} ; O' _{k ( i , j )} = O _{k ( i , j )} + D; this way is to equalize the image and The original image is regarded as a linear difference relationship, that is, the difference between the intensity values of the composite gray scale image and the first gray scale image is regarded as a correction value D, and after the correction value is obtained, the original image is added to the original image to obtain The hue is corrected for the equalized image O' _{( i , j )} and the result is shown in Fig. 4B.

The above description shows how to enhance the image contrast after using the histogram method and the like, and still make the hue not shift, and obtain an equalized image that is stronger than the original image but has a hue similar to the original image.

Please refer to FIG. 5A to FIG. 5C, and FIG. 5A is a color image diagram of the retinal image. Figure 5B is a color image of the first equalized image of Figure 5A. Figure 5C shows a color image of the second equalized image of Figure 5A. In pictures 5A to 5C, the retinal image is used as an illustration of the image enhancement method of the present invention, the purpose of which is to highlight the fine blood vessel image in the retinal image, but still preserve the hue of the original retinal image to solve the increase of the contrasted image. Distortion problem. In Figure 5A, which is not processed by the image enhancement method 100, the blood vessels in the retinal image are less pronounced. Figures 5B and 5C show the processing using the image enhancement method 100, using the ratio R and the correction value D, respectively. The result of image enhancement. It can be seen that not only the blood vessels are highlighted, but the hue is not offset, and a relatively correct contrast retinal blood vessel image can be obtained.

The selection of the above two correction methods depends on the characteristics of the original image and the purpose of image enhancement. There is no good relationship between the two. In other image enhancement methods that are not illustrated, the above-mentioned correction values may also use other modes, and are not limited to the ratios or differences disclosed in the present invention. In principle, as long as the source of the correction value takes into account the composite grayscale image and the intensity value of the first grayscale image, the operational relationship between the two intensity values in the correction value and the corresponding correction method of the equalized image can be reasonable The way to implement it.

With the above-described embodiments, the present invention has at least the following advantages: First, a Gabor filter is used to generate a second grayscale image, Instead of the conventional technique, the histogram equalization is directly performed on the original grayscale image, and the hue of the original image can be preserved. Second, the combination of the filtered image and the original grayscale image does not involve the destruction of the original color image, and at the same time, better image detail performance can be obtained. Third, the generation of the correction value is open. According to the target of the image processing, the correction value can be adjusted by itself, and the calculation procedures of various correction values are simple, and the equalized image can be obtained relatively quickly.

Although the present invention has been disclosed in the above embodiments, it is not intended to limit the present invention, and the present invention can be modified and modified without departing from the spirit and scope of the present invention. The scope is subject to the definition of the scope of the patent application attached.

100‧‧‧Image enhancement method

110~170‧‧‧Steps

Claims (5)

An image enhancement method includes: reading an original image and converting the original image into a first grayscale image; using a Gabor filter to enhance the first grayscale image to form a second grayscale Combining the first grayscale image and the second grayscale image to form a composite grayscale image; obtaining one image parameter of the composite grayscale image; using the image parameter to perform a histogram of the composite grayscale image And calibrating the composite gray scale image after the histogram equalization step and the intensity value of the first gray scale image to obtain a correction value, wherein the correction value is the composite gray scale image and the first gray a ratio or difference of intensity values of the order image; and enhancing the original image by the correction value to obtain a first-order image. The image enhancement method of claim 1, wherein the image parameter comprises a probability density function and a cumulative distribution function. The image enhancement method of claim 1, wherein the Gabor filter enhances the first grayscale image by a calculation program to form the second grayscale image, the calculation program is: Where u = xcos θ + ysin θ, v = - xsin θ + ycos θ, θ is the direction of the Gabor filter, σ _u and σ _v are the standard deviations of the Gaussian function on the u-axis and the v-axis, respectively, and ω is the modulation frequency. The image enhancement method of claim 1, further comprising: multiplying the hue value of the original image by the correction value to obtain a hue value of the equalized image. The image enhancement method of claim 1, further comprising: adding a hue value of the original image to the correction value to obtain a hue value of the equalized image.