基于图像信息熵统计直方图的图像增强算法

目的为了解决图像因亮度较大造成的成像效果不佳、局部细节不清楚等问题。方法将直方图均衡化技术(Histogram Equalization, HE)引入图像信息熵域,提出对比度弱化的图像信息熵统计直方图自适应均衡化算法(Contrast-reduced Adaptive Entropy Histogram Equalization, CRAEHE)。以各个灰度级信息熵统计值为基础,先将原图像分割成若干个子区域,对每个子区域的灰度信息熵统计值进行阈值截取,补充到子区域内各个灰度级上,再对子区域进行信息熵直方图均衡化处理。采用USC-SIPI和CBSD432数据集图像,用图像灰度均值、标准差、平均梯度、信息熵等参数对实验样本进行质量评价。结果文中算法处理结果较原图灰度均值下降了7.94%,标准差平均提高了52.22%,信息熵平均提高了19.86%,平均梯度提高了57.19%。结论文中算法增强了选自数据集里的过亮图像的细节,并使图像整体细节与质量都得到了改善,该算法的处理结果较其他处理实验样本的主观质量提升明显,对光照强度适应范围广。

Image Enhancement Algorithm Based on Entropy Statistical Histogram of Image Information

The work aims to solve the problem of poor imaging quality and unclear local details due to the large brightness of the image. The histogram equalization (HE) was introduced into the image information entropy domain to propose the contrast-reduced adaptive entropy histogram equalization (CRAEHE). Based on the statistical information entropy of each gray level, the original image was first divided into several sub-regions, and the statistical gray-scale information entropy of each sub-region was intercepted by thresholds and supplemented onto each gray level in the sub-region. Then, the sub-region performed the information entropy histogram equalization processing. With USC-SIPI and CBSD432 dataset images, the quality of the experimental samples was evaluated by such parameters as image gray average, standard deviation, average gradient and information entropy. The gray average of the original image processed by the proposed algorithm was decreased by 7.94%, the standard deviation was increased by 52.22% averagely, the information entropy was increased by 19.86%, and the average gradient was increased by 57.19%. The proposed algorithm enhances the details of over-lighted images selected from the datasets, and improves the overall image details and quality. The processing results of this algorithm are significantly improved compared to the subjective quality of other processed experimental samples, and the range of adaptation to light intensity is wide.