應用於低複雜度區塊運動估計之位元轉換技術研究

Study on Bit Transform Techniques for Low Complexity Block Motion Estimation

楊友勝(You-Sheng Yang)

指導教授 ： 陳柏宏

國立虎尾科技大學/電機資訊學院/電子工程系碩士班/碩士(2014年)

摘要

本論文提出一種應用於視訊編碼區塊運動估計的新型二位元轉換技術。最先提出的二位元轉換技術係計算個別門檻視窗的期望值與標準差，爾後以此兩值計算出兩個位元平面的轉換閥值。而之後提出的二位元轉換技術係利用個別門檻視窗的二階導數濾波運算結果，與期望值搭配計算出兩個位元平面的轉換閥值。此外, 使用二階導數的二位元轉換提出了一種加權型非匹配點數量的相似度判斷準則，減少在原先二位元轉換技術中出現局部最少非匹配點數的機會。然而，使用門檻視窗期望值計算位元轉換閥值，在區塊運動估計時可能會找不到最相似區塊，因此本論文藉由改變原始二位元轉換之第一位元平面轉換方式，以原始一位元轉換技術的轉換方式取代，並採用加權計算的相似度判斷準則。實驗結果顯示，本論文所提出的新型二位元轉換技術相較於相關的二位元轉換技術，不僅大部分影像序列能夠獲得較好的編碼影像品質，亦能維持相近的運算複雜度。

關鍵字

視訊編碼 ； 位元轉換 ； 區塊運動估計

並列摘要

This thesis proposes a novel two-bit transform for block motion estimation of video coding. The first two-bit transform technique computes bit transform thresholds for two bit planes by using mean and standard deviation of threshold window. The second two-bit transform technique uses the second order derivative of threshold window and threshold window mean to compute bit transform threshold for two bit planes. However, it is very likely to find a non-best match block for block motion estimation by using threshold window mean to compute bit transform threshold. In addition, the second two-bit transform proposed a weighted NNMP(number of non-matching point) matching criterion to reduce the chance to derive a local minimum NNMP in the first two-bit transform. Hence, the thesis generates the first bit plane by using bit transform threshold computed with Multiband-pass filtered image used in one-bit transform, and adopts weighted NNMP. Experimental results shows the proposed novel two-bit transform technique not only achieves better video quality than the previous two-bit transforms, while also maintains similar computational complexity.

並列關鍵字

Video coding ； Bit transform ； Block based motion estimation

參考文獻

[1] Charles Poynton, 2012, Digital Video and HD, Second Edition, Morgan Kaufmann.

[2] Paul Read, Mark-Paul Meyer, 2000, Restoration of Motion Picture Film, First Edition, Butterworth-Heinemann.

[3] Iain Richardson, 2002, Video Codec Design: Developing Image and Video Compression Systems, First Edition, Wiley.

[4] Arch Luther, Andrew Inglis, 1999, Video Engineering, Third Edition, McGraw-Hill Professional.

[7] S. Cui, Y. Wang, and J.E. Fowler, 2002, “Mesh-based motion estimation and compensation in the wavelet domain using a redundant transform”, in IEEE ICIP., Rochester, pp. 693-696, September.