基于Contourlet的图像感知质量评价

Image Perceptual Quality Assessment Using Contourlet Transform

图像感知质量评价是图像信息工程的基础技术之一.结合人类视觉系统(HVS)的感知特性,在Con-tourlet变换域建立了一个新的可计算JND门限模型,该模型综合考虑了HVS的空间频率敏感性、方向敏感性、对比度掩盖与邻域掩盖特性.由于邻域掩盖模型的引入,能够有效鉴别图像中平滑、边缘与纹理结构区域对失真的不同掩盖强度,实现更加精确的掩盖阈值计算.借助于建立的JND模型,定义每个系数的感知误差,进而建立了感知质量评价标准HVSNR.实验结果表明该定量评价标准能够有效匹配人类的视觉感觉.

Image perceptual quality assessment is a key problem in image processing engineering.According to the perceptual characters of Human Visual System（HVS）,a Just Noticeable Distortion（JND） threshold model is constructed using contourlet transform,which can well qualify spatial frequency sensitivity,orientation sensitivity,contrast masking and neighborhood masking effects of HVS.As a result of taking account of neighborhood masking additionally,this JND model can distinguish the different masking intensity of smoothness,edge and texture domain,and implement more accurate JND threshold.Based on our JND threshold model,the perceptual error between contourlet coefficients is defined,and then a quantitative perceptual quality metric HVSNR is proposed.Experiments demonstrate that our metric can provide quality evaluation well correlated with those given by human observers.