遞迴式最佳像素預估法應用於

視訊壓縮之錯誤追蹤

摘要

最新發展出的視訊標準，如H.264，不論在時間域或空間域都使用複雜的預測方式，以使其壓縮效率增加，但一旦發生錯誤仍可能會造成嚴重的錯誤漫延效應。因此本論文便在於探討將遞迴式最佳像素預估法(ROPE)進一步結合應用於同步式逆向錯誤追蹤(SBET)之中，以達到防制錯誤漫延與增加強健性的目標。

若編碼端的狀態能和解碼端的狀態同步，則錯誤蔓延的效應就能完全被制止，因此我們假設有回授通道存且在編碼端可由某些方式知道解碼端所使用的錯誤隱藏機制，讓編碼端得知解碼端的狀態，即可藉由同步式逆向錯誤追蹤將解碼端狀態重建於編碼端，來達到錯誤漫延的防制。然而，在往復延遲時間內的情況則是無法預知的，因此為防範可能的錯誤蔓延，我們提出將遞迴式最佳像素預估法與其進一步結合應用，讓編碼端用遞迴的方式計算出解碼端像素失真量的期望值，此失真量會包含量化失真、錯誤蔓延的錯誤和錯誤隱藏所帶來的錯誤。這些計算都會整合到碼率失真模型裡，以便視訊編碼器在有通道錯誤的環境下選出最佳的編碼模式。從實驗結果來看，結合遞迴式最佳像素預估法與同步式逆向錯誤追蹤較原始遞迴式最佳像素預估法有顯著之改進，可以讓視訊編碼器有更好的抗錯性。

關鍵字 – 錯誤隱藏、錯誤追蹤、遞迴式最佳像素預估、回授通道。

Recursive Optimal per-Pixel Estimate Applied to

Error Tracking in Video Compression

Abstract

The most recent H.264 video coding utilizes complex predictions in both the temporal and spatial domains to achieve higher compression efficiency.Unquestionably, such predictions may cause serious error propagation effects when suffering from transmission errors. Therefore, the objective of this thesis is to combine the Recursive Optimal per-Pixel Estimate (ROPE) with the Synchronous Backward Error Tracking (SBET) algorithm for terminating the error propagation and enhancing error robustness.

If the state of the encoder can synchronize to that of the decoder, the error propagation effects can be entirely terminated. Therefore, we assume that a feedback channel is available and the encoder can be aware of the decoder’s error concealment by external means. The SBET is utilized to reconstruct the state of decoder in the encoder for the prevention of error propagation. However, it is impossible to foresee the decoder state within the round-trip delay time. Consequently, to prevent such possible error effects, we further enhance SBET by ROPE. The encoder will recursively keep estimating the expected decoder pixel distortion. The distortion of frame reconstruction at the decoding side includes the quantization error, propagation error, and concealment error. These estimates are integrated into the rate-distortion model for optimal mode decision of the encoder in the error-prone environment. From simulation results, we can find that the joint SBET and ROPE algorithm shows much better error resilience improvement than ROPE.

Keywords - error concealment, error tracking, Recursive Optimal per-Pixel Estimate, feedback channel.