Combination of color information and properties of human visual system was used to assess the quality of color images.Structural information of a color image was described by quaternion.The three RGB channels of a color image were encoded into the three imaginary parts of a quaternion.The real part of a quaternion was the local variance of the luminance layer of the color image and the effect of the detail structure on the image quality was emphasized accordingly.An algorithm to calculate the singular value decomposition(SVD) of a quaternion matrix using its equivalent complex matrix was used to perform the decomposition of the image.The angle between singular vectors of the reference image and the distorted image was used to measure the structural similarity of the two images.Result from experiments shows that the proposed method is better consistency with human visual system characteristics than MSE

PSNR and SSIM.It can also be used to assess the images whose size is different from that of the reference image.