The CIE color difference formula is not suitable for predicting color difference between complex images. In order to quantify color differences between the original and reproduction, an image difference algorithm based on correlation of image attribute was put forward. The formula used the Contrast Sensitivity Functions (CSF) to reduce the spatial complexity. The CSF was used to remove information that is imperceptible to the visual system. The algorithm determined the color attribute weights by means of correlation analysis between image pairs. Psychophysical experiments confirmed that the metric correlates strongly with human perception. It was concluded that the algorithm can be used to assess the performance of color image compression, color image enhancement algorithms, gamut mapping algorithms and so on.