Abstract
The signal-to-noise ratio, SNR, is well used for evaluating compressed images as an objective evaluation value. However, the impressions of observing images, even if the SNRs are equal, differ greatly, and it is known that it is dependent on images. Objective evaluation measures based on the lightness difference and the color difference between an original and a compressed images were used for the purpose of creating the measure which correspond subjective evaluation values and do not dependent on images. Furthermore, the average lightness and the average color differences as the average of colorimetric values of some pixels in consideration of the blur at the observation of images were also used. The four images in SCID were compressed by using JPEG and HPK. Comparison observation between the compressed and the original images was carried out, and subjective evaluation values were obtained. Correlations with the subjective evaluation value and the SNR, the lightness difference, the color difference, the average lightness difference and the average color difference between compressed images and original images were calculated. Consequently, it was found that average color difference showed the highest correlation coefficient to subjective evaluation value.
