On a perceptual rate-distortion model for color image coding

The Mean Square Error (MSE) is a common distortion measure used to assess image quality although it is not always justified when compared to the human observer. In this work we present a Rate-Distortion approach to color image compression based on subband transforms using perceptual optimization of the compression quality. This approach is based on minimization of the Weighted Mean Square Error (WMSE) of the encoded image, which better corresponds to quality assessment of the human eye. Based on the new approach, new optimized compression algorithms are introduced using the Discrete Cosine Transform and the Discrete Wavelet Transform. We compare the new algorithms to presently available algorithms such as JPEG. Our conclusion is that the new WMSE optimization approach offers superior performance when a human observer is considered.