No-reference image quality assessment for compressed images based on DCT coefficient distribution and PSNR estimation

We propose a new no-reference image quality assessment model nPSNR (No-reference PSNR) for JPEG compressed images. The model performs in DCT domain and the DCT coefficient distribution is used. This method estimates the MSQE (mean-squared quantization error) of a decoded image with the distributions of AC coefficients and DC coefficients of the encoded image, for PSNR estimation in DCT domain without reference. We test the proposed model on the selected images from the TID2008 database. Then, the computational scores (nPSNR) are compared with the ground truth values (MOS) based on three commonly-used performance criteria. Experimental results demonstrate that the proposed approach is more consistent with the subjective perception than the state-of-the-art full-reference image quality assessment schemes.