基于视觉掩盖效应和奇异值分解的图像质量评测方法.pdfVIP

第l6卷 第4期 光学 精密工程 VoI．16 NO．4 2008年 4月 Optics and Precision Engineering Apr．2008 文章编号 1004—924X(2008)04—0706—08 基于视觉掩盖效应和奇异值分解的图像质量评测方法 袁 飞，黄联芬，姚 彦 (厦门大学信息科学与技术学院通信工程系，福建厦门361OO5) 摘要：将人眼视觉特性和图像本征特性相结合，设计了一套基于视觉掩盖和奇异值分解的图像质量评测方法。根据人眼 的视觉敏感性函数(CSF)将参考图像和待测图像做噪声掩盖处理，剔除视觉敏感区外的图像信息；然后，根据人眼视觉 特性提取参考图像的纹理区和平坦区，并作为待测图像检测区的划分依据，分别计算参考图像和待测图像在各自纹理区 和平坦区的奇异值；最后，通过奇异值测度得到待测图像在纹理区和平坦区的质量劣化程度。实验表明：通过纹理区和 平坦区的区分检测，可提高检测敏感度并据此判断劣化及影响程度。同时，检测结果的主观相似性(CC)可达到0．896， 均方根误差(RMSE)为5．58，较PSNR和SSIM具有更好的性能。该方法实现简单，便于系统集成，可在涉及视频及图 像质量检测的领域得到较好应用。 关 键 词：图像质量；奇异值分解；视觉掩盖；客观评价 中图分类号：TP391．4 文献标识码：A Image quality evaluation based on visual masking ar value decomposition YUAN Fei，HUANG Lian—fen，YAO Yan (Communication Engineering Department of Information and Engineering Institute，Xiamen University，Xiamen 361005，China) Abstract：A image quality evaluation method based on visual masking effect and singular value~decom— position is proposed．According to the Contrast Sensibility Function(CSF)，the original image and the processed image are preprocessed to remove the signals outside the sensible region．On the basis of the property of the Human Visual System (HVS)，the texture and plat regions of the original image are extracted to used as a division standard．Then，the singular values of the original and processed ima— ges are calculated both in texture and plat regions．Finally，by measuring the difference between the singular value metrics of the original and the processed images，the image quality can be evaluated． The experimental results show that this method is sensible to the errors perceived by human—being and can improve discrimination for the distortion region．The correlation between the result and t