Evaluating Image Quality of DIBR-synthesized Views in Texture and Depth Domains

Abstract

Accurately predicting the quality of depth-image-based-rendering (DIBR) synthesized images is of great significance in promoting DIBR techniques. Recently, Many DIBR-synthesized image quality assessment (IQA) algorithms based on quantifying the distortion existed in texture images have been proposed. However, these methods ignore the damage of DIBR algorithms on the depth structure of DIBR-synthesized images and thus fail to evaluate the visual quality of DIBR-synthesized images accurately. To this end, we present a DIBR-synthesized image quality assessment metric with multi-modal information (i.e. integrating texture and depth information), dubbed as MMI. MMI predicts the quality of DIBR-synthesized images by jointly measuring the texture structure and depth structure of the synthesized image. The design principle of our MMI is that the local geometric distortion, introduced by DIBR techniques in the hole-filling process, destroys the texture structure and the depth structure of DIBR-synthesized images. Thus, we can accurately evaluate DIBR-synthesized image quality by a joint representation of texture structure and depth structure. Experiments show that our MMI is better than the competing state-of-the-art IQA algorithms in predicting DIBR-synthesized image quality.