视频图像的邻域统计直方图均衡化增强算法

Neighborhood Calculation Histogram Equalization Enhancement Algorithm for Video Image

针对视频监控图像相邻帧之间场景变化小的特点,提出一种适用于视频图像的邻域统计直方图均衡化算法来提升图像的对比度。根据相邻视频帧的场景相关性,对图像当前帧进行邻域信息统计,利用Laplace算子得到前一帧图像中的背景和细节信息,从而选择合适的增强参数,生成视频图像的优化直方图。使用优化后直方图的均值点将其分割成2个部分得到各自的变换函数,对视频图像进行直方图均衡化操作。实验结果表明,该算法能够提升视频图像的对比度,保持图像的平均亮度,减弱过增强现象,保留视频中兴趣区域的细节信息。可以获得的最小绝对平均亮度误差值为1.4741,最大熵值为7.0993,最大峰值信噪比值为20.6710。

Considering the adjacent frames scene small changes of surveillance video images, this paper proposes a neighborhood calculation histogram equalization algorithm for video image to enhance the contrast. This algorithm employs the adjacent video frame correlation, calculates the neighborhood information of the frame, chooses a suitable enhancement parameter according to the former frame background and detail information obtained by Laplace operator, and generates a modified histogram of the image. The algorithm separates the modified histogram into two parts using its mean value to get two transformation functions. Histogram equalization is operated for the latter frame. Experimental results show that this algorithm efficiently enhances the contrast of video images,preserves the mean brightness,alleviates the over-enhancement problem and keeps the interest region details. Compared with the traditional algorithms,it acquires the lowest average absolute error of brightness value as 1. 474 1,highest entropy as 7. 099 3,and highest peak signal to noise ratio as 20. 671 0 .