|
基於編碼增益下H.264解碼器移動補償之複雜度控制 摘
要 視訊壓縮編碼與網路的蓬勃發展促成了多樣化的多媒體應用，如錄影與視訊會議。即時視訊編解碼應用在行動裝置上也已經相當地普遍。最新的視訊壓縮標準H.264提供了許多編碼工具來達到高編碼效率，相對在編解碼端也增加較高的運算複雜度。然而在行動裝置上即時視訊解碼運算複雜度是有限的，因此控制解碼器的運算複雜度使其低於複雜度限制並且維持最佳位元率-失真效能是非常重要的課題。 本論文針對H.264解碼器之移動補償運算，提出一個利用編碼增益階層(Coding-gain-based
layer, CGL)複雜度控制機制來控制解碼器複雜度。實驗結果顯示，一般影像採用所提出之編碼增益階層三(CGL-3)進行編碼所產生之串流，最高可節省達37%之解碼時間且PSNR只稍降0.44dB。而本機制也可以針對解碼器不同複雜度的限制下，提供相符之編碼串流來進行即時解碼，利用此機制產生之串流所需解碼複雜度和實際上相比，平均誤差僅為2.43
%，亦即在可接受之視訊品質需求且解碼器複雜度有限狀況下，本研究所提之複雜度控制機制可以有效的控制解碼端的複雜度，達到複雜度控制之目的。 關鍵字：視訊壓縮、複雜度控制、移動補償、編碼增益 |
|
Coding-Gain-Based
complexity control for motion compensation in H.264 video decoder Video Applications in mobile devices
become more and more popular. The latest video compression standard H.264/AVC
provides various coding tools to achieve high coding efficiency at the
expense of high computational complexity. Because the computation capability
of a mobile device is generally constrained, the full-scale H.264 video
decoding may be not allowed for a mobile device. Therefore, a complexity
control mechanism which adjusts the complexity of video coding computational
complexity and maintains the Rate-Distortion (RD) performance is important. Most studies on complexity control
focus on the encoding side. However, we propose a complexity control
mechanism for video decoders because decoders are more popularly used. Motion
compensation (MC) is the most complexity-consuming operation in H.264 video
decoding, the complexity control of MC is critical and the first one to be
considered in our work. This research proposes a Coding-Gain-Based layer
(CGL) mechanism which controls MC complexity of the decoder by controlling
allowable search point locations and partition modes in the encoder. The
computational overhead of the proposed mechanism is totally just in the
encoder. The simulation results show that the proposed mechanism can reduce
decoding time up to 37% with less than 0.44 dB video quality degradation
compared with no complexity constrained case. It can efficiently control the
decoding complexity with only 2.43% error rate on average. Keywords:
video coding, complexity control, motion compensation, coding gain |