A wavelet approach to compressed image quality measurement

The understanding and characterization of various forms of image degradation due to compression is important. The most commonly used image distortion measure is the mean squared error (MSE). However the MSE value does not provide a good subjective performance measure, since it does not take human perception into account. We present a wavelet approach to measure the visual quality of compressed images based on a psycho-visual human vision model. Wavelet filter banks are shown to have a good space frequency localization property and can be directly linked to the Michelson contrast and compactly supported features. Experimental results are provided to demonstrate the effectiveness of the Haar wavelet.