Computational Model for Stereoscopic Image Quality Prediction

Abstract

In the modern era of Internet along with 3D imaging and communication system, many user-end applications require the estimation of quality of 3D images directly from the bit streams, as the original image may not be available. Though several metrics have been proposed in literature to assess the full reference perceptual quality of 3D images, however no reference quality assessment is still undeveloped which is a challenging issue. Therefore, in this paper, we propose a no reference stereoscopic image quality evaluation model based on image artifacts and disparity measure with the incorporation of Human visual system (HVS) characteristics. Based on HVS, we believe that perceptual artifacts of any image are strongly dependent on local features, such as plane and non-plane areas. For this reason, plane and non-plane area based blockiness and blur artifacts and also disparity are measured in this model. The experimental results show that the proposed model gives high correlation with subjective Mean Opinion Score (MOS).