Xiaoquan Yi

Improved frame-layer rate control for H.264 using MAD ratio

In recent years, rate control plays an increasing important role in real-time video communication applications using MPEG-4 AVC/H.264. An important step in many existing rate control algorithms, which employs the quadratic rate-distortion (R-D) model, is to determine the target bits for each P frame. This paper aims in improving video distortion, due to high motions or scene changes, by more accurately predicting frame complexity using the statistics of previously encoded frames. We use mean absolute difference (MAD) ratio as a measure for global frame encoding complexity. Bit budget is allocated to frames according to their MAD ratio, combined with the bits computed based on their buffer status. Simulation results show that the H.264 coder, using our proposed algorithm with virtually little computational complexity added, effectively alleviates PSNR surges and sharp drops for frames caused by high motions or scene changes.

Fast pixel-based video scene change detection

This paper proposes a new simple and efficient method to detect abrupt scene change based on only pixel values. Conventional pixel-based techniques can produce a significant number of false detections and missed detections when high motion and brightness variations are present in the video. To increase scene change detection accuracy yet maintain a low computational complexity, a two-phase strategy is developed. Frames are first tested against the mean absolute frame differences (MAFD) via a relaxed threshold, which rejects about 90% of the non-scene change frames. The rest 10% of the frames are then normalized via a histogram equalization process. A top-down approach is applied to refine the decision via four features: MAFD and three other features based on normalized pixel values - signed difference of mean absolute frame difference (SDMAFD*), absolute difference of frame variance (ADFV*), and mean absolute frame differences (MAFD*). Experimental results show that our method contributes to higher detection rate and lower missed detection rate while maintaining a low computational complexity, which is attractive for real-time video applications.

Improved H.264 rate control by enhanced MAD-based frame complexity prediction

Abstract This paper presents a revised rate control scheme based on an improved frame complexity measure. Rate control adopted by both MPEG-4 VM18 and H.264/AVC use a quadratic rate–distortion (R–D) model that determines quantization parameters (QPs). Classical quadratic R–D model is suitable for MPEG-4 but it performs poorly for H.264/AVC because one of the important parameters, mean absolute difference (MAD), is predicted through a linear model, whereas the MAD used in MPEG-4 VM18 is the actual MAD. Inaccurately predicted MAD results in wrong QP and consequently degrades rate–distortion optimization (RDO) performance in H.264. To overcome the limitation of the existing rate control schemes, we introduce an enhanced linear model for predicting MAD, utilizing some knowledge of current frame complexity. Moreover, we propose a more accurate frame complexity measure, namely, normalized MAD, to replace the current use of MAD parameter. Normalized MAD has a stronger correlation with optimally allocated bits than that of the predicted MAD. To minimize video quality variations, we also propose a novel long-term QP limiter (LTQPL). Finally, a dynamic bit allocation scheme among basic units is implemented. Extensive simulation results show that our method, with inexpensive computational complexity added, improves the average peak signal-to-noise ratio (PSNR) and reduces video quality variations considerably.

Context Adaptive Lagrange Multiplier (CALM) for Rate-Distortion Optimal Motion Estimation in Video Coding

In this paper, we propose an efficient and practical algorithm to dynamically adapt the Lagrange multipliers for each macroblock based on the context of the neighboring or upper layer blocks to improve rate-distortion performance. Our method improves the accuracy for the detection of true motion vectors as well as the most efficient encoding modes for luma, which are used for deriving the motion vectors, and modes for chroma. Simulation results for H.264/advanced video coding video demonstrate that our method reduces bit rate significantly and achieves peak signal-to-noise ratio gain over those of the joint model (JM) software for all sequences tested, with negligible extra computational cost. The improvement is particularly significant for high motion high-resolution videos. This paper describes our work that led to our Joint Video Team adopted contribution (included in software JM 12.0 onward), collectively known as context adaptive Lagrange multiplier (CALM).

Improved Normalized Partial Distortion Search With Dual-Halfway-Stop for Rapid Block Motion Estimation

Motion estimation is a critical yet computationally intensive task for video encoding. In this paper, we present an enhancement over a normalized partial distortion search (NPDS) algorithm to further reduce block matching motion estimation complexity while retaining video fidelity. The novelty of our algorithm is that, in addition to the halfway-stop technique in NPDS, a dual-halfway-stop (DHS) method, which is based on a dynamic threshold, is proposed, so that block matching is not performed against all matching candidates. An adaptive search range (ASR) mechanism based on inter block distortion further constrains the searching process. Simulation results show that the proposed algorithm has a remarkable computational speedup when compared to that of full search and NPDS algorithms. Particularly, it requires less computation by 92-99% and encounters an average of only 0.08 dB PSNR video degradation when compared to that of full search. The speedup is also very significant when compared to that of fast motion estimation algorithms. This paper describes our work that led to our joint video team (JVT) adopted contribution (included in software JM 10.1 onwards) as well as later enhancements, collectively known as simplified and unified multi-hexagon search (SUMH), a simplified fast motion estimation.

Rapid block-matching motion estimation using modified diamond search algorithm

Due to the considerable computational complexity of full-search (ITS) in motion estimation, many suboptimal but fast block-matching algorithms (BMAs) have been developed. Among them, the diamond search (DS) series is the most promising method. To further reduce complexity and improve performance, we propose a modified diamond-search (MODS) algorithm for rapid block matching based on the well-known DS algorithm. A novel fine granularity halfway-stop (FGHS) method based on a dynamic block distortion threshold is also proposed. To avoid being trapped in local optima, unlike some small DS methods, MODS adaptively starts with a relatively large search pattern for high motion blocks which are automatically determined via the first block matching distortion. The threshold is obtained via a linear model utilizing already computed distortion statistics. Experiments show that the proposed algorithm achieves less search points with no significant PSNR degradation when compared to that of FS and other fast BMAs.

Rate control using enhanced frame complexity measure for H.264 video

Bit rate control is an important issue for wireless and Internet video streaming. The paper presents a revised rate control scheme based on an improved frame complexity measure. Rate control adopted by both MPEG-4 VM18 and H.264/AVC uses a quadratic rate-distortion (R-D) model that determines quantization parameters (QPs). The classical quadratic R-D model is suitable for MPEG-4, but it performs poorly for H.264/AVC because one of the important parameters, mean absolute difference (MAD), is predicted through a linear model, whereas the MAD used in MPEG-4 VM18 is the actual MAD. Inaccurately predicted MAD results in the wrong QP and consequently degrades rate distortion optimization (RDO) performance in H.264/AVC. To overcome the limitation of the existing rate control schemes, we introduce an enhanced linear model for predicting MAD, utilizing some knowledge of current frame complexity. Moreover, we propose a more accurate frame complexity measure, namely, normalized MAD, to replace the current use of MAD parameters. Normalized MAD has a stronger correlation with optimally allocated bits than that of the predicted MAD. Finally, a dynamic bit allocation scheme among basic units is implemented. Extensive simulation results show that our method, with inexpensive added computational complexity, improves the average peak signal-to-noise ratio (PSNR) considerably, by up to 1.2 dB, and reduces PSNR variances significantly, by up to 63%.

On enhancing H.264 rate control by PSNR-based frame complexity estimation

This work presents an enhancement to the frame layer buffer status-based H.264 rate control method. The enhancement is by using a PSNR-based frame complexity estimation to improve the existing mean absolute difference based (MAD-based) complexity measure. Bit allocation to each frame is not just computed by encoder buffer status but also adjusted by a combined frame complexity measure. Simulation results show that the H.264 encoder, using the proposed algorithm, can achieve better visual performance than that of the existing H.264 JVT rate control method (JVT/spl I.bar/G012).

Improved partial distortion search algorithm for rapid block motion estimation via dual-halfway-stop

Block motion estimation is a critical, yet computationally intensive, task for video encoding. Many fast block matching algorithms have been developed. Partial distortion search (PDS) algorithms generally produce less video quality degradation of the predicted images than those of conventional fast block matching algorithms (BMAs). However, the speedup gain of PDS algorithms is usually limited. We present an enhancement over a normalized PDS (NPDS) algorithm to reduce block matching motion estimation complexity further and improve video fidelity. The novelty of our algorithm is that, in addition to the halfway-stop technique in NPDS, a dual-halfway-stop (DHS) method, which is based on a dynamic threshold, is proposed so that block matching is not performed against all searching points. The dynamic threshold is obtained via a linear model utilizing already computed distortion statistics. An adaptive search range mechanism based on inter block distortion further constrains the searching process. Simulation results show that the proposed algorithm has a remarkable computational speedup. Particularly, it requires 92.0-99.4% less computation than full search (FS) and 8.0-91.0% less than NPDS. It encounters an average of 0.07 dB video degradation in PSNR performance compared to FS whereas it gains 0.01-0.12 dB over the NPDS algorithm.

Bit rate distribution analysis for motion estimation in H.264

This paper analyzes bit rate distribution of full search and fast block matching motion estimation algorithms in H.264. It shows that a well-designed fast block matching algorithm not only can speed up motion estimation process, but can also improve rate-distortion performance. This is in contrast to the common belief that full search always yields optimal rate-distortion performance, in comparison to all fast BMAs.