Haar wavelet decomposition based blockiness detector and picture quality assessment method for JPEG images

We propose a new method for blockiness distortions identification and quality assessment of JPEG images in the absence of any reference images using Haar wavelet decomposition. Blockiness artifacts contaminating JPEG encoded images can be modeled as a particular edge structure that transforms into another when subjected to edge detection algorithm. Some features that give meaningful information for blockiness distortions and quality assessment are subsequently derived from edge detected image by means of 3-level Haar wavelet transform. The validity of our quality assessment technique was tested against publicly available images quality assessment database whilst the proposed blockiness detector was evaluated subjectively using our own subjective test. Our experiments show that the proposed quality metric and blockiness distortions detector demonstrate satisfying performances.