基于最大熵模型的双直方图均衡算法

为改进亮度保持双直方图均衡算法的不足，提出基于最大熵模型的动态范围优化方法，扩展了双直方图均衡算法的应用范围，使之不仅适用于正常亮度图像，对低照度及高亮图像也能取得较好的效果.算法首先选用大津法确定直方图数据分割点；然后对初始直方图进行预处理；根据所提出的最大熵模型确定最佳的动态范围分割点；最后进行双直方图均衡得到增强图像.本文选取多个图像数据库进行测试，并与BBHE（Brightness preserving Bi-Histogram Equalization）、BPCLBHE（Brightness Preserving and Contrast Limited Bi-Histogram Equalization）、ESIHE（Exposure based Sub Image Histogram Equalization）和DRSHE（Dynamic Range Separate Histogram Equalization）进行比较，同时将信息熵、对比度和NIQE（Natural Image Quality Evaluator）作为客观评价指标.实验结果证明，本文算法对各类图像均具有较好的主观视觉效果和客观评价指标，在保留细节的同时兼顾了对比度的增强.

Maximum Entropy Model Based Bi-histogram Equalization Algorithm

In order to improve the drawback of brightness preserving bi-histogram equalization (BHE) algorithms,a dynamic range optimization method based on maximum entropy model is proposed to extend the application range of BHE algorithms,which makes it not only suitable for normal brightness image,but also can get good effect on low illumination and high brightness image.Firstly,the segmentation point decided by Otsu divides the original histogram into two sub-histograms.And then,a hybrid adjustable method is proposed to pre-process the initial histogram.After that,based on the proposed maximum entropy model,the best dynamic range segmentation point is determined by an ergodic optimization method.Finally,the BHE procedure outputs the final contrast enhanced image.Multiple image databases are selected for testing and compared with BBHE (Brightness preserving Bi-Histogram Equalization),BPCLBHE (Brightness Preserving and Contrast Limited Bi-Histogram Equalization),ESIHE (Exposure based Sub Image Histogram Equalization) and DRSHE (Dynamic Range Separate Histogram Equalization),and we use the entropy,contrast and NIQE (Natural Image Quality Evaluator) as objective evaluation indexes.The experimental results demonstrate that the proposed algorithm outperforms other state-of-the-art BHE algorithms on almost all types of images,with better subjective visual effective and better objective score of quality evaluation,and it takes the enhancement of contrast into account while preserving details.