一种基于粒子群优化的高斯混合灰度图像增强算法  

Gaussian mixture grayscale image enhancement algorithm based on particle swarm optimization

作　　者：陈皓月[1] 钱钧[1] 姜文涛[1] 杨一洲[1] 宋磊[1] 黄西莹[1] 

机构地区：[1]西安应用光学研究所,陕西西安710065

出　　处：《应用光学》2017年第4期592-598,共7页Journal of Applied Optics

基　　金：国家国际科技合作专项资助(2015DFA10670);陕西省科技厅资助项目(2015KJXX-62)

摘　　要：提出一种采用粒子群优化(PSO)的高斯混合灰度图像增强算法。该算法首先采用高斯混合模型(GMM)对输入图像的灰度直方图建模,并采用模型中高斯成分的有效交点来分割直方图。随后,该算法将每个直方图区间的灰度值转换到合适的输出区间,生成增强后的灰度图像,其中转换函数由输入直方图区间的高斯成分和累积分布经过粒子群优化后的参数决定。实验结果显示,该方法生成的图像视觉效果较好,对原图像和纹理细节丰富图像分别进行图像增强,增强后的图像信息熵分别是4.746 6和7.952 6,灰度平均梯度为6.970 6和37.386 1。A Gaussian mixture model（GMM） based grayscale image enhancement algorithm using particle swarm optimization（PSO） is proposed. The algorithm uses GMM to build a model for gray level histogram of the input image, and uses the significant interaction points of Gaussian components in the model to partition the histogram into a certain amount of sub intervals. Then, according to mapping function, the gray value in each interval is transformed to appropri- ate output interval. The enhanced image is generated by transforming output interval data ac- cording to PSO optimized parameters. Experimental results show that image visual effects gen- erated by Gaussian mixture gray image enhancement algorithm with PSO are better. After image processing of original image and texture rich image, the information entropy of enhanced image is 4. 746 6 and 7. 952 6 respectively, the average gray gradient is 6. 970 6 and 37. 386 1.

关 键 词：图像处理 灰度图像增强 高斯混合模型 粒子群优化 

分 类 号：TN911.73[电子电信—通信与信息系统]