中文摘要：

对于视频编码等视频应用来说，准确客观地评价其视频质量至关重要．在客观结构相似度模型的基础上，结合人眼视觉特性提出了一种新的视频质量评价方法一基于区域加权的结构相似度视频质量评价方法．该方法基于三维Sobel矩阵提取的视频时空特性，将视频帧定义为强边缘、弱边缘和平滑三大区域进行加权求得质量分，并结合了人眼对失真的边缘块和失真帧较为敏感等视觉特性．在LIVE测试数据库中的实验结果表明，该方法与人类视觉的主观评价结果具有较好的一致性．

英文摘要：

To evaluate the video quality accurately and objectively is important for video application, especially for video coding. Based on Structure Similarity （SSIM） model, we propose a new video quality assessment method by combining the human visual characteristic with the video quality assessment using a region-weighted model. Based on the video spatio-temporal characteristic extracted using three-dimensional Sobel matrix, the proposed method divides the video frame into strong edge region, weak edge region and smooth region, by which the frame-level scores are determined. Considering human visual characteristics such as sensitivity to the distortion block and distortion frame, the proposed method incorporates the frame-level quality scores into sequence-level quality score. The proposed method is tested on the LIVE subjective quality video database, and the results show the better performance using the proposed method in matching subjective ratings when comparing other classical methods.