基于图像结构信息复数表示与奇异值分解的灰度图像质量评价方法

Image quality assessment based on complex number representation of image structure and singular value decomposition

为了进一步突出图像结构中人眼敏感的重要特征,采用复数矩阵表示图像结构,将图像的局部方差和像素灰度值分别作为复数的实部和虚部。进而对复数矩阵进行分块奇异值分解,分析了传统奇异值分解图像质量评价方法的特点,将复数矩阵每一分块奇异值分布的标准差作为分块图像结构的表征,分别计算参考图像与待测图像对应图像分块奇异值标准差,从而得到了图像结构失真映射图谱,通过计算图谱中的数据分布特征得到最终的量化评价结果。采用LIVE数据库中包含5种失真类型的779幅测试图像验证所提的算法。试验结果表明,本文方法采用复数矩阵描述图像结构信息,平衡了对各种失真类型的敏感程度,与人眼视觉感知(HVS)的一致性优于传统方法。

In order to accentuate the complicated information that human eyes are sensitive to in an image,complex matrix is used to describe image structure. Local variance and pixel value are taken as the real part and imaginary, respectively. Singular value decomposition is performed on each block of the complex matrix, but standard deviation of the singular value distribution is used as the description of the image block. The standard deviation is calculated for the reference image and distorted image, respectively. The matrix composed of standard deviations corresponding to each block is used to obtain the distor tion map. Quantified results are obtained by calculating the data distribution of the map. 779 distorted images in the LIVE database are used to test the performance of the proposed method. Experimental results show that the performance improvement is achieved by using complex matrix to describe image structure, which balances the distortion sensitivity. Consistency with the human visual perception of the proposed method is better than that of traditional methods.