基于对比度受限直方图均衡化的水下海参图像增强方法

针对水下图像受到水下复杂光照的影响导致图像对比度差的现象,采用对比度受限自适应直方图均衡化方法(contrast-limited adaptive histogram equalization,CLAHE)对水下海参图像进行增强处理,算法首先将原始图像分割成若干个子区域并且大小相同,再选取特定值对每个子区域的直方图进行截取,并将截取下的像素均匀分配到每个灰度级,最终得到限定对比度直方图。并通过研究算法中的相关参数,得到适用于水下海参图像增强的参数值,取得了更好的增强效果。通过评价函数均方差(mean squared error,MSE),峰值信噪比(peak signal to noise rate,PSNR)和信息熵(information entropy)对比CLAHE方法和其他一些方法,结果显示CLAHE算法在水下海参图像提高质量和保持图像细节方面表现出更好的性能,为以后水下机器人的识别定位提供了方便。

Method for image intensification of underwater sea cucumber based on contrast limited adaptive histogram equalization

Because of the suspended solids in the underwater environment, and light absorption and scattering, underwater sea cucumber image has the weaknesses of illumination uneven, low contrast, various kinds of noise and soon, causing difficulties for the identification for underwater robots.In this paper, image enhancement technology for underwater sea cucumber image was studied, and a method called Contrast limited Adaptive Histogram Equalization(CLAHE) was proposed to deal with the underwater sea cucumber images.We used sea cucumbers which grew up in Shandong Haiyang Thousand Island Lake aquaculture base as the experimental subject, and recorded video by artificial dive underwater with the digital camera(Canon Power Shot G12) in July 2015, in order to get all kinds of images(single sea cucumber, sea cucumber with substrates, many sea cucumbers) filtrated the pictures from all the videos, at last we got about 200 images which contained a variety of circumstances.The image enhancement algorithm used in the article has the following steps: Firstly, the original image was divided into several sub regions of the same size(each sub region was continuous and non overlapping).Secondly, we selected a specific value to make sure that the number of pixels in each gray level was no more than this value, and then used the specific value to intercept the histogram of each sub region, and the intercepted pixels were evenly distributed to each gray level.Thirdly, we made histogram equalization to the gray histogram of each sub area after shearing.Fourthly, bilinear interpolation was used to get gray value of the central point of each sub block, and taking these points as reference points, the mapping of each pixel point in the image was determined by the mapping of by the four reference points around.Finally, the enhancement of the underwater sea cucumber images was finished by using the method of Contrast limited Adaptive Histogram Equalization, and we also used other image processing methods (such as: histogram equalization, linear conversion) dealing with sea cucumber images.Then through subjective judgment(observe the change of processed image and its histogram, compare the changes and find out which method is the best one), we found out that images processed by HE had the shortcoming of noise over enhancement, processed by Linear conversion turn up color distortion.Also by evaluation functions: Mean squared error(MSE), Peak signal to noise rate(PSNR), and information entropy were used for objectively evaluated the method used in this paper and the other methods.We got the average value of MSE and PSNR, the information entropy and processing time by processing 200 images, and it turned out that the value of the method used in this paper was better than the other methods: the average of MSE was about 29.570 5, PSNR was about 24.119 4, and information entropy was about 6.936 4.Generally speaking, the value of MSE was smaller, the result was better, and the value of PSNR and information entropy was bigger, the result was better.The study showed that CLAHE took great advantages of several other methods to achieve good results.To get better parameters which are suitable for underwater sea cucumber image, we also improved the algorithm by studying the related parameters, while the split window was too large, the strengthen will be weaken; too small window will cause over enhancement, and found that the effect was best when the window size was about 32×32.The experimental results show that: CLAHE algorithm shows better performance in improving the quality of underwater sea cucumber image and maintaining the details of the images than the other methods, the value of objective evaluation function has a better promotion, and all of this provides convenience for the identification of underwater robot positioning.