No abstract

No abstract

No abstract

No abstract

A new video coding tool, sample adaptive offset (SAO), is introduced in this paper. SAO has been adopted into the Working Draft of the new video coding standard, HighEfficiency Video Coding (HEVC). The SAO is located after deblocking in the video coding loop. The concept of SAO is to classify reconstructed pixels into different categories and then reduce the distortion by simply adding an offset for each category of pixels. The pixel intensity and edge properties are used for pixel classification. To further improve the coding efficiency, a picture can be divided into regions for localization of offset parameters. Simulation results show that SAO can achieve on average 2% bit rate reduction and up to 6% bit rate reduction. The run time increases for encoders and decoders are only 2%.

No abstract

This paper proposes a novel multi-class hybrid-boost learning algorithm for multi-pose face detection and facial expression recognition. This system detects human face in different sizes, various poses, partial-occlusion, and different expressions. The contribution of this paper is the hybrid boosting algorithm combining the Haar-like (local) features and Gabor-like (global) features. The experimental results show that our system has better performance than the others.

In this paper, a one-pass encoding algorithm is proposed for adaptive loop filter (ALF) in high-efficiency video coding (HEVC). ALF can improve both subjective and objective video quality, but it also requires a lot of encoding passes (i.e. picture buffer accesses) that will significantly increase external memory access, encoding latency, and power consumption. Therefore, we propose a method to estimate filtering distortion without performing real filter operation. The number of encoding passes can be effectively reduced from 16 to 1. Combined with an initial guess of filter-on/off blocks by using time-delayed filters, the proposed one-pass algorithm only induces average 0.17% BD-rate increase.