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  

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

作　　者：张发军 彭文刚[2] 虞成俊 何孔德 ZHANG Fa-jun;PENG Wen-gang;YU Cheng-jun;HE Kong-de(Key Laboratory of Robot and Intelligent System,Yichang 443002,China;College of Mechanical and Power Engineering of Three Gorges University,Yichang 443002,China)

机构地区：[1]机器人与智能系统市重点实验室,宜昌443002 [2]三峡大学机械动力学院,宜昌443002

出　　处：《科学技术与工程》2020年第21期8629-8636,共8页Science Technology and Engineering

基　　金：国家自然科学基金(51275273);机器人与智能系统市重点实验室资助项目(2019JX07)。

摘　　要：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。In an engineering application of wood texture recognition,wood image of poor contrast could be produced because of uneven lighting and other factors,which may affect the accurate recognition of wood texture.Therefore,an algorithm of contrast-limited adaptive histogram equalization(CLAHE)was applied to deal with wood images of poor contrast.Firstly,the original image was grey processed on weighted average grayscale and then divided into several sub-regions of equal and continuous size.Secondly,an appropriate threshold was chosen to intercept the histogram of each sub-regions,and distribute intercepted pixels to other grayscales.Finally,the pixels in the processed image were interpolated after using bilinear interpolation to obtain a histogram-equalized image with contrast limitation and bilateral filter was also used to process the equalized image for noise reduction and edge preservation of wood image.Six hundreds of wood images were shared,of which 450 were used for training set and 150 for verification set.Evaluation functions including mean absolute error(MAE)and peak signal to noise rate(PSNR)were adopted.To further verify the superiority of CLAHE,an experiment of recognition of wood texture was performed.The results show that the CLAHE method is effective in processing wood texture image.The results build a theoretical basis for accurate recognition of wood texture.

关 键 词：木板纹路识别 图像增强 对比度受限自适应直方图均衡化算法 双边滤波器 直方图均衡化 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]