An Efficient Mode Selection Prior to the Actual Encoding for H.264/AVC Encoder

Abstract

Abstract-Many video compression algorithms require decisions to be made to select between different coding modes. In the case of H.264, this includes decisions about whether or not motion compensation is used, and the block size to be used for motion compensation. It has been proposed that constrained optimization techniques, such as the method of Lagrange multipliers, can be used to trade off between the quality of the compressed video and the bit rate generated. In this paper, we show that in many cases of practical interest, very similar results can be achieved with much simpler optimizations. Mode selection by simply minimizing the distortion with motion vectors and header information produces very similar performance to the full constrained optimization, while it reduces the mode selection and over all encoding time by 31% and 12%, respectively. The proposed approach can be applied together with fast motion search algorithms and the mode filtering algorithms for further speed up.

A novel motion classification based intermode selection strategy for HEVC performance improvement

“…We measure the respective change between current block and co-located blocks of different frames and by regulating this change we can produce the motioncompensated block in the reference frame [25]- [27]. The Phase Matched Error (PME) is obtained by subtracting the motioncompensated reference block from the current block.…”

“…Therefore, our motivation is to trim down the computational time of HEVC by smart selection of appropriate ME and MC modes with the exploration of motion based on a number of features in the videos. For more appropriate decision of ME and MC modes, in this paper, we incorporate a new technique-HEVC-PC (HEVC with phase correlation) that approximates relative displacement information of the current block against the reference block [25][26] to predict the motion type of a CU. The motion based CUs are then encoded by the modes in the higher coding depth levels while for the CUs without motion are encoded by the modes in lower coding depth levels.…”

Sensor-Assisted Video Encoding for Mobile Devices in Real-World Environments

“…According to [32], the methods for reducing the video encoding complexity for H.264/AVC can be classified into three categories: 1) reducing the number of search points in motion estimation, e.g., the fast motion estimation and predictive motion estimation we mentioned in Sections II-A1 and II-A2; 2) reducing the number of candidate encoding modes for a given macroblock, e.g., [33]- [35]; and 3) applying efficient mode selection by improving the optimization cost function, e.g., [32] itself proposed a simplified Lagrange multipliers cost function for reducing the number of operations in mode selection. A particular benefit of the third category, as claimed by [32], is that any reduction achieved by this category can also be applied on the top of the other two categories.…”

“…A particular benefit of the third category, as claimed by [32], is that any reduction achieved by this category can also be applied on the top of the other two categories. Our proposed SaVE scheme falls into the first category so it can be combined with other low-power video encoding schemes, e.g., [32], to achieve further complexity reductions.…”

Fast Bi-Directional Prediction Selection in H.264/MPEG-4 AVC Temporal Scalable Video Coding

“…Based on the non-zero quantized transform coefficients, the proposed schemes [19], [20] construct an accurate RD estimator to avoid calculating the true RD values during the mode decision process. Furthermore, a modified Langrangian cost function [21], composed of the distortion, the motions, and the required header, eliminates entirely the entropy coding in the decision process. Another popular approach in fast mode selection is the so-called early termination [22]- [28].…”