A Quality Assessment of Watermarked Color Image Based on Vision Masking

The quality assessment of the watermarked color image is one of the most important aspects that evaluate the performance of watermarking system. PSNR is the common quality assessment of the watermarked image. But the results measured by PSNR usually do not match to human vision well and result in misleading. The discrete cosine transform was used as the image analysis tools. The psychophysical research -- the luminance, texture, frequency and color masking on human vision were analyzed and used in masking strength calculation. Then a vision masking calculation method of image DCT coefficients was given in this paper. Based on the masking strength calculation, a new quality assessment of watermarked color image was proposed. Experimental results showed that the new quality assessment outperforms PSNR. The distorted image quality values calculated using the new quality assessment are consistent with the perception of human vision and have high correlation and consistency with subjective values. proposed a DCT(Discrete Cosine Transform) domain watermark assessment method based on Laplacian model. For the same watermarking algorithm, embed location and watermark sequence, the evaluation results become worse with the increase of the embed strength. The algorithm can't distinguish the difference when we changed the length of embedded sequence under the same embed strength. By using the frequency masking, Ming, Hao and Yang (10) proposed a weighted image measurement, while the luminance and texture masking were not considered. By calculating the just noticeable difference (JND) for each DCT coefficient of image with watson's JND model, Yang and Liang (11) assigned