为了克服现有的直方图均衡算法在图像细节信息增强的同时无法保持输入图像亮度的缺点,以部分重叠局部直方图均衡算法(POSHE)为基础,将递归分解的思想引入子块图像直方图均衡过程,提出了一种亮度保持的部分重叠局部直方图均衡算法。子块图像根据其累积概率密度(CDF)的中值,递归地进行不同深度的分解;然后,对每一个分解的子图像在其灰度范围内进行直方图均衡,并将均衡后的子图像合并;最后,根据移动步长对子块图像进行线性或双线性插值消除块效应对图像细节信息的影响。实验结果表明,这种改进的部分重叠局部直方图均衡算法的峰值信噪比(PSNR)比传统方法平均提高了0.063~6.633 dB。在实际应用中,该算法既能有效地增强图像的细节信息,又能保持输入图像的亮度,并且彻底地消除了块效应对图像均衡结果的影响,使均衡后的图像具有更加自然的视觉效果。

The traditional histogram equalization is unable to preserve the input image brightness,when the contrast of an image is enhanced. Based on a Partially Overlapped Sub-block Histogram Equalization algorithm (POSHE),the recursive separateness method is used in the process of sub-block histogram equalization. A brightness preservation partially overlapped local histogram equalization algorithm which can preserve the brightness is proposed. The sub-block image is separated recursively into different sub-images with the cumulative density function (CDF). Then,each sub-image histogram is equalized within its respective gray range,and all the sub-images are incorporated into the equalized sub-blocks. Both linear and bilinear interpolations are used in the equalized sub-block based on a step size to avoide the generation of blocking effects and the influence on image details. Experimental results show that the average Peak Signal Noise Ratio(PSNR) is improved by 0.063~6.633 dB compared with those of traditional interpolation methods. In fact,the local image details are enhanced,and the input image brightness is also preserved effectively. For the influence of the blocking effects on the image is eliminated,the equalized image shows a more natural visual impression.