The paper deals with prospective 3D video transmission systems that would use compression of both multiview video and depth maps. The paper addresses the problem of quality of views synthesized from other views transmitted together with depth information. For the state-of-the-art depth map estimation and view synthesize techniques, the paper proves that AVC/SVC-based multiview video coding technique can be used for compression of both view pictures and depth maps. The paper reports extensive experiments where synthesized video quality has been estimated by use of both PSNR index and subjective assessment. Defined is the critical value of depth quantization parameter as a function of the reference view quantization parameter. For smaller depth map quantization parameters, depth map compression has negligible influence on fidelity of synthesized views.
Paper concerns accumulation of video-quality losses in multiple HEVC encoding and decoding cycles. The results of extensive experiments with standard video test sequences are reported. General conclusions has been drawn about relative quality losses after each cycle, as well as about total quality loss with respect to first cycle. In general, quality losses after each cycle are decreasing and total quality loss is saturating after several cycles. Total quality losses after several cycles are hardly dependent on sequence content.
This paper presents a detailed, quantitative analysis of video quality losses in a homogeneous HEVC video transcoder together with the analysis of the origin of those quality losses and the influence of quantization step alignment on transcoding. With the use of HM15.0 reference software and a set of test video sequences, the cascaded pixel domain video transcoder (CPDT) concept has been used to gather all the necessary data needed for the analysis. This experiment was performed for a wide range of source and target bitrates. The essential result of the work is an extensive evaluation of CPDT, commonly used as a reference in works on effective video transcoding. So far, no such extensively performed study has been made available. The quality degradation between a transcoded video and a video that would be the result of direct compression of the original video at the same bitrate as the transcoded one has been reported. The dependences between quality degradation caused by transcoding and bitrate changes of the transcoded data stream are clearly presented in graphs.
Nowadays, research focusing on robotics, autonomous vehicles, and scene analysis shows a clear need for the ability to accurately reconstruct three-dimensional environments. One of the methods allowing to conduct such a reconstruction is to use a set of cameras and image processing techniques. This is a passive method. Despite being, in general, less accurate than its active counterparts, it offers significant advantages in numerous applications in which active systems cannot be deployed due to limited performance. This paper provides a theoretical analysis of the accuracy of estimating 3D positions of objects present at a given scene, based on images from a set of cameras. The analysis assumes a known geometrical configuration of the camera system. The important limiting factor in the considered scenario is the physical resolution of sensors - especially in the case of systems that are supposed to work in real time, with a high FPS rate, as the use of high-resolution cameras is difficult in such circumstances. In the paper, the influence of the geometric arrangement of the cameras is studied and important conclusions about the potential of three-camera configurations are drawn. The analysis performed and the formulas derived help predict the boundary accuracy values of any system using a digital camera. The results of an experiment that confirm the theoretical conclusions are presented as well.
This paper presents a novel approach to occlusion handling problem in depth estimation using three views. A solution based on modification of similarity cost function in optimization algorithm (on example of graph cuts) is proposed. During depth estimation via optimization algorithms like graph cuts similarity cost function is constantly updated so that only non-occluded pixels in the side views are considered. For the side views, virtual depth maps are synthesized and occluded regions are detected. Basing on that, similarity cost function is updated for correspondence search only in non-occluded regions of the side views. The experimental results, performed with use of a well-known 3D video test sequences, show that the proposed approach in application for virtual view synthesis for next generation of 3D-television, provides gains of about 1.25 dB of PSNR related to the state-of-the-art technique implemented in MPEG Depth Estimation Reference Software.