This thesis presents an improved Color Image Quality Metric to evaluate the quality of color images. The Metric uses two color spaces that resemble the human visual system: lαβ and Lab. The color images are quantized in different channels and then ranked by all subjects through a series of experiments. We analyzed the results by using nonparametric statistics: the trends of single-channel distorted images were discussed and followed by the comparison of the mixed single-channel distorted images. Finally, we compared the calculating results between the metric in the literature and what we proposed based on mixed dual-channel distorted images. The comparisons of lαβ and Lab color space was carried out as well.