Fast encoder decision for texture coding in 3D-HEVC

2015 18th International Conference on Computer and Information Technology (ICCIT), 2015

§Abstract-Multiview video employs the utilization of both texture and depth video information from different angles to create a 3D video for more realistic view of a scene. Unlike texture, depth video is a gray scale map that represents the distance between the camera and 3D points in a scene. Existing multiview video coding (MVC) techniques including 3D-High Efficiency Video Coding (HEVC) standard encode both texture and depth videos jointly by exploiting texture video information for the corresponding depth video coding (DVC) to reduce computational time as the texture and depth videos have motion similarity in representing the same scene. This strategy has two limitations: (i) more bits and computational time might be required due to the large residuals for the misalignment between depth and texture edges and (ii) switching between different views may require more times due to the increased dependency between texture and depth. In this paper, we propose an independent DVC technique using HEVC (a video coding standard for single view) so that we can improve the rate distortion (RD) performance and reduce computational time by improving switching speed. For this, we use motion features to reduce a number of motion estimation (ME) and motion compensation (MC) modes in HEVC. As we use motion feature which is the underlying criteria for selecting different modes in the standard and then we select a subset of modes which can provide almost the same RD performance. Experimental outcomes reveal a reduction of 48% encoding time of HEVC encoder with similar RD performance and better interactivity.

IET Image Processing, 2017

High efficiency video coding (HEVC) achieves higher coding efficiency than previous standards but introduces a large computational complexity. Because HEVC adopted some new advanced tools and the most prominent one is the flexible hierarchical coding structures which include coding unit (CU), prediction unit (PU), transform unit (TU). All of those units must be test through rate-distortion optimisation. Since the CU is highly content dependent it is not efficient to test all the modes. In this study, the authors propose a fast coding algorithm for HEVC based on video contents. The authors statistically analysis the features of video contents from three aspects, the pixel gradient, the block mean value, and the block variance of CU. Then, jointly use these features with the CU depth levels and the prediction modes of spatiotemporal adjacent CUs to realise fast CU depth level decision and fast prediction mode decision. The experimental results show that the proposed algorithm can save 58.9 and 57.6% computational complexity on average with only average 1.8 and 1.9% bitrate losses under 'random access, main' and 'low-delay, main' conditions, respectively.

2015

HEVC implements a candidate vector list for merge and skip modes. The construction of this list has been extensively studied in the JCT-VC group (see for instance JCTVC-G039). It has been shown in JCTVC-I0293 that it is possible to improve the HEVC coding performance by adding in the merge list copies of the first candidate shifted by an arbitrary offset. The same basis is considered in this document and applied to disparity compensation. A gain of 0.3 % is obtained on average on side views. 1

2017 25th European Signal Processing Conference (EUSIPCO), 2017

The 3D extension of the High Efficiency Video Coding (HEVC) standard achieves large compression rates thanks to the addition of several tools to encode multiview and depth information on top of those available in HEVC. The use of such tools incur in a very large computational demand, which can be a serious problem in power and computationally-constrained devices and applications. However, not all information contained in an image is fundamental to the viewer, so that different levels of computational effort can be employed when encoding different image regions. The Region of Interest (ROI) concept is used in this work to classify each Coding Unit (CU) as foreground, heterogeneous background and homogeneous background. Then, a simplified encoding process is employed in those regions classified as homogeneous background, terminating earlier the partitioning process in texture CUs, while still still performing the regular decisions in areas classified as ROI. Experimental results show ...

International Conference on Advances in Pattern Recognition, 2015

In this paper, a parametric texture based video coding scheme is proposed for multi-view videos. The proposed scheme exploits the perceptual redundancy of the textured region in a statistical manner which is ignored in the multi-view video coding scheme of H.264/MVC. This statistical nature of texture is independent of the 3D scene structure. The set of most similar texture macroblocks over different frames is termed as a motion thread. In the proposed scheme, each P slice/frame of multi-view video is categorized into non-texture blocks which are encoded by H.264/MVC and texture blocks which are encoded by our novel approach. These texture based motion threads are encoded with the help of Spatio-Temporal Autoregressive (STAR) model. These motion threads covers the temporal and interview movement of macroblocks maintaining consistency while STAR model exploits the temporal and interview perceptual redundancy in neighborhood. The proposed scheme achieves average gain as high as 4.9 dB PSNR over H.264/MVC for given multi-view videos.

IFIP Advances in Information and Communication Technology, 2013

High-Efficiency Video Coding is currently proposed as the newest and most efficient video coding standard by the ITU-T Video Coding Experts Group and the ISO/IEC Moving Picture Experts Group. Compression improvement relative to existing standards is estimated in the range of 50%. It is a block-based hybrid video coding algorithm that introduces several novel features compared to MPEG-4 like Coding Units associated with Prediction Units and Transform Units, Advanced Motion Vector Prediction, minimization of a rate-distortion Lagrangian cost, directional orientations for intra-picture prediction etc. The core algorithm of Context-Adaptive Binary Arithmetic Coding is based upon that of the MPEG-4 standard. View synthesis algorithms for HEVC stereo and 3D encoding are expected to be finalized as a standard extension. A Multi-view Video Coding scheme based upon the estimation of correlated parameter sets of elastic models between views is wherein adopted. The order of the tensor equals the number of multiple views. Underlying distributions are updated step-by-step. They are modeled according to context indices.

In addition to the texture, multiview video employees the utilization of depth coding for the reconstruction of 3D video and Free viewpoint video. Standing on some texture-depth correlations, a number of methods in literature reuses texture motion vector for the corresponding depth coding to reduce encoding time by avoiding costly motion estimation process. However, texture similarity metric is not always equivalent to the corresponding depth similarity metric especially at edge levels. Since their approaches could not explicitly detect and encode acute edge motions of depth objects, eventually, could not reach the similar or improved rate-distortion (RD) performance against High Efficiency Video Coding (HEVC) reference test model (HM). With a view to more accurate motion detection and modeling, the proposed technique exploits an extra Pattern Mode comprising a group of pattern templates (GPTs) with different rectangular and non-rectangular object shapes and edges compared to the existing HEVC block partitioning modes. Selecting an extra mode i.e., the Pattern mode, it can avoid further exploration of modes at next depth levels, thereby reducing encoding time by avoiding further branching. Moreover, the proposed Pattern Mode only encodes the motion areas and skips the background areas. The experimental results show that the proposed technique could save 30% encoding time and improve average 0.1dB Bjontegard Delta peak signal-to-noise ratio (BD-PSNR) compared to the HM.

IEEE Transactions on Circuits and Systems for Video Technology, 2014

Advanced multiview video systems are able to generate intermediate viewpoints of a 3D scene. To enable low complexity free view generation, texture and its associated depth are used as input data for each viewpoint. To improve the coding efficiency of such content, view synthesis prediction (VSP) is proposed to further reduce inter-view redundancy in addition to traditional disparity compensated prediction (DCP). This paper describes and analyzes rate-distortion optimized VSP designs, which were adopted in the 3D extensions of both AVC and HEVC. In particular, we propose a novel backward-VSP scheme using a derived disparity vector, as well as efficient signaling methods in the context of AVC and HEVC. Additionally, we put forward a novel depth-assisted motion vector prediction method to optimize the coding efficiency. A thorough analysis of coding performance is provided using different VSP schemes and configurations. Experimental results demonstrate average bit rate reductions of 2.5% and 1.2% in AVC and HEVC coding frameworks, respectively, with up to 23.1% bit rate reduction for dependent views.

2016

HEVC video coding standard uses a structure of flexible Coding Unit (CU) quad tree. This flexibility increases the coding efficiency and allows the standard to achieve better rate-distortion performance compared to its prior H.264/MPEG-4 AVC. The improvement of coding efficiency came at the cost of high encoding complexity. This paper presents a fast algorithm to reduce the encoding complexity by an early termination decision with reference frame selection. Experimental results for seven standard video sequences show that the proposed algorithm can save up to 53.79% of the encoding time compared to the original HM14. Different video sequences with different frame rates, quantization parameters and resolutions were used in the experiments. The paper also presented the effect of changing resolutions, frame rates and bit rates on compression efficiency for both HM and the proposed algorithm. In comparison with the original software, the proposed algorithm presents significant improveme...

2012 Picture Coding Symposium, 2012

Motion and disparity estimations are employed in Multi-view Video Coding (MVC) to remove redundancies present between temporal and different viewpoint frames, respectively, in both the color and the depth multi-view videos. These constitute the major computational expensive tasks of the video encoder, as iterative search for the optimal mode and its appropriate compensation vectors is employed to reduce the Rate-Distortion Optimization (RDO) cost function. This paper proposes a solution to limit the number of modes that are tested for RDO to encode the interview predicted views. The decision is based on the encoded information obtained from the corresponding Macro-block in the Base view, identified accurately by using the multi-view geometry together with the depth data. Results show that this geometric technique manages to reduce about 70% of the estimation's computational time and can also be used with fast geometric estimations to reduce up to 95% of the original encoding time. These gains are obtained with little degradation on the multi-view video quality for both color and depth MVC.