对比度受限下直方图均衡化处理的木板纹路图像增强算法

在木板纹路识别的工程应用中,光照不均匀等因素会产生对比度差的木板图像,将影响木纹的准确识别,针对此现象采用改进的对比度受限自适应直方图均衡化方法(contrast-limited adaptive histogram equalization,CLAHE)对木板纹路图像进行增强处理,首先借助加权平均灰度化对原始图像进行灰度处理,再利用CLAHE算法将灰度图像分割成若干个大小相等连续的子块,选取合适的阈值对每个子块的直方图进行截取,将超过阈值的像素均分到其他的灰度级中,使用双线性插值对图像进行插值运算,得到对比度受限下直方图均衡化图像,最后使用双边滤波器(bilateral filter)对均衡化图像进行降噪保边处理。共用600幅木板图像,其中450幅图像作为训练集图像,150幅作为验证集图像,通过平均绝对误差(mean absolute error,MAE),峰值信噪比(peak signal to noise rate,PSNR)以及最终实验结论对比CLAHE算法和其他一些常规算法,工程运行结果表明:CLAHE方法处理木板纹理图像具有较好的运行效果,为木板纹理的准确识别提供了理论依据。

Algorithm for Image Enhancement of Wood Texture Based on Contrast-Limited Adaptive Histogram Equalization

In the engineering application of wood texture recognition, wood image with poor contrast can be produced because of uneven lighting and other factors, and it will affect the accurate recognition of wood texture. In this paper, an algorithm called contrast-limited adaptive histogram equalization (CLAHE) is applied to deal with wood image with poor contrast. Firstly, the original image is processed by using a method of weighted average gray level, and then divided into several sub-regions of equal size and continuous. Secondly, choosing appropriate threshold to intercept the histogram of each sub-regions, and distributing intercepted pixels to other gray levels. Finally, the pixels in the processed image are interpolated after using bilinear interpolation to obtain a histogram equalized image with contrast limitation and bilateral filter is also used to process the equalized image for noise reduction and edge preservation of wood image. The paper shares 600 wood images, 450 of which are used as training set images and 150 as verification set images. Evaluation functions including Mean absolute error (MAE) and Peak signal to noise rate (PSNR) are adopted. To further verify the superiority of CLAHE, we do an experiment of recognition of wood texture, and conclusion of final experiment show that the CLAHE method is effective in processing wood texture image, it also provides a theoretical basis for the accurate identification of wood texture.