Depth Image Quality Assessment for View Synthesis Based on Weighted Edge Similarity

Leida Li, Xi Chen, Yu Zhou, Jinjian Wu, Guangming Shi; Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, 2019, pp. 17-25

Abstract


With the increasing prevalence of multi-view and freeview displays, virtual view synthesis has been extensively researched. In view synthesis, texture and depth images are typically fed into a depth-image-based-rendering (DIBR) algorithm to generate the new viewpoints. In contrast to the enormous amount of research effort on the quality assessment of texture images and rendering process, much less effort has been dedicated to the quality evaluation of depth images. To fill this gap, this paper presents a quality metric of depth images for view synthesis. Depth image represents information relating to the distance of the surfaces of scene objects from a viewpoint, and edge conveys key location information in depth image, which is extremely important in view rendering. Therefore, the proposed metric is developed with emphasis on measuring the edge characteristics of depth images. Firstly, a similarity map is computed between the distorted and reference depth images by combining intensity and gradient information. Then an adaptive weighting map is calculated by integrating depth distance and location characteristics in the depth image. Finally, an edge indication map is computed and utilized to guide the pooling process, producing the overall depth quality score. Extensive experiments and comparisons on the public MCL-3D database demonstrate that the proposed metric outperforms the relevant state-of-the-art quality metrics.

Related Material


[pdf]
[bibtex]
@InProceedings{Li_2019_CVPR_Workshops,
author = {Li, Leida and Chen, Xi and Zhou, Yu and Wu, Jinjian and Shi, Guangming},
title = {Depth Image Quality Assessment for View Synthesis Based on Weighted Edge Similarity},
booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops},
month = {June},
year = {2019}
}