影像預測編碼中動態位元數，像素位元數與計算機效能的分析

Abstract

預測編碼是動態影像壓縮的重要過程。無論Intra或Inter預測編碼，都從畫面中的鄰近區塊找尋與目前區塊最近似的區域，互相比對後將結果編碼成描述區塊動態的動態位元與描述區塊內像素大小的像素位元。找尋再消除區塊之間的累贅，使得動態位元數與像素位元數總和少於預測前的位元總數，這就是預測編碼提升壓縮率的方法。 不同的預測編碼方式產生不同的動態位元與像素位元，而兩者的位元數總合直接影響到壓縮率。不同的預測編碼方式也需要不同的計算機運算次數。因此三個要素決定了預測編碼是否成功：動態位元數、像素位元數、與計算機運算次數。

如何在最少的計算機運算次數下，產生最少的動態位元數與像素位元數是預測編碼的終極目標。本文以H.263，H.264與MPEG4為研究對象，模擬三個視訊壓縮標準的Intra與Inter預測編碼表現。無論Intra或Inter，模擬結果皆顯示H.264在高昂的運算次數代價下得到最少的動態位元數與像素位元數，也就是最高的壓縮率。H.263的壓縮率最低，但是計算機運算次數最少。MPEG4的表現居中。預測編碼時相同的搜尋區域，H.264的計算機運算次數是H.263的七倍。

動態位元與像素位元的編碼方式直接影響到壓縮率高低。本文並且介紹、比較、歸納三個壓縮標準的編碼方式。

Prediction Coding plays an important role in Motion Image Compression. When coding current macroblock, Intra and Inter Prediction Coder search for the nearest neighbor macroblock, compensate it, and encode it into Motion Bits and Residue Bits, which express the motion and pixel value of current macroblock. Prediction Coder search for and eliminates the redundancies between macroblocks, thus decreases the amount of Motion Bits and Residue Bits to increase the total compress ratio.

The number of Motion and Residue Bits effect compress ratio, and searching costs Computing Power. We may say that Motion Bits, Residue Bits, and Computing Power are the key factors of Prediction Coding.

The ultimate goal of Prediction Coding is to encode macroblock to fewest Motion and Residue Bits under lowest Computing Power cost. The thesis simulates Prediction Coding by H.263, H.264, and MPEG4. The simulation result shows that H.264 brings the fewest bits under highest Computing Power cost. H.263 brings the lowest compress ratio with lowest Computing Power cost. The thesis states that during prediction coding with same searching area, the computing time of H.264 is seven times than H.263.

Codeword of Motion and Residue Bits directly effects the compress ratio. The thesis also discusses and compares the codeword design in the three standards.