高动态范围图像增强算法  

HIGHLY DYNAMIC RANGE IMAGE ENHANCEMENT ALGORITHM

作　　者：席志红[1] 赵蓝飞[1] 

机构地区：[1]哈尔滨工程大学信息与通信工程学院,黑龙江哈尔滨150001

出　　处：《计算机应用与软件》2014年第5期202-205,255,共5页Computer Applications and Software

基　　金：国家自然科学基金项目(60875025)

摘　　要：针对显示设备无法还原高动态范围图像的真实效果,提出一种基于分层色阶映射的图像增强算法。该算法首先通过Retinex模型与快速双边滤波算法估计光照图像。然后利用高斯混合模型和期望最大算法将光照图像概率分布分割为多层高斯模型混合而成。并通过迭代获得各高斯概率层对应的最优校正系数。最后叠加各层并获得增强后的光照图像。实验证明,该算法在整体上实现了图像动态范围的压缩,图像增强效果明显。Due to display devices cannot restore the true effects of highly dynamic range images,the article puts forward an image enhancement algorithm based on hierarchical tone mapping. Firstly the algorithm applies Retinex model and fast bilateral filtering algorithm to estimate the illumination images. Secondly it applies Gaussian mixture model and expectation maximization algorithm to split the illumination image probability into hierarchical Gaussian models mixed together. Thirdly by iteration it acquires the optimal gamma coefficients corresponding to each Gaussian probability hierarchies. At last the hierarchies are layered to obtain the enhanced illumination image. Experiment verifies that the algorithm realizes the compression of image dynamic range while its image enhancement effects are obvious.

关 键 词：高动态范围 RETINEX 双边滤波 高斯混合模型 期望最大 色阶映射 动态系数 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]