Energy Loss Estimation Based Reference-Free Quality Assessment of DIBR-Synthesized Views

The processes of warping and rendering in Depth Image-Based Rendering (DIBR) are usually performed without a reference image. Thus, an effective and fast reference-free (RF) image quality assessment (IQA) model devised for DIBR-synthesized views is more favorable. To this aim, this research propose a novel RF IQA method of DIBR-synthesized views based on Energy Loss Estimation. Firstly, considering that the geometric distortions caused by the processes of warping and rendering might increase the high-frequency (HF) information of a DIBR-synthesized view, the influence of the geometric distortion on the visual quality of a synthesized view can be effectively evaluated by estimating the HF information energies of a given DIBR-synthesized view in the log-gabor domain. Secondly, we estimate the energy loss in the process of JPEG compression as the image complexity, which is used to overcome the problem of different amounts of HF information contained in different content images. Finally, the proposed RF DIBR-synthesized IQA metric is obtained by using the energy loss in JPEG compression to normalize the HF information energy in the log-gabor domain. Experiments conducted on two public datasets demonstrate that the proposed method is advantageous over the relevant state-of-the-art RF IQA metrics.