H.264之方向導向錯誤隱藏及預測式邊界相稱演算法

No Thumbnail Available

Date

2005

Journal Title

Journal ISSN

Volume Title

Publisher

Abstract

雜訊在資料的傳輸中是一個很重要的議題。在編碼系統中,有許多程序亦是為了抵抗雜訊而製作。H.264在其規格中,訂制一個可變區塊序列編碼(Flexible Macroblock Order;FMO)的機制,當資料流中的某一個部份發生錯誤時,可利用在影像中周圍的區塊資訊進行補償動作,以達到更好的補償效果。本研究提出一個以方向導向的錯誤隱藏機制,在H.264的CODEC中,配合FMO之機制,利用前一張畫面中共位區塊(collocated block)之移動向量的方向特性,選擇一個較適合的候選區塊進行錯誤隱藏的工作。 另一方面,傳統的邊界相似度演算法(Boundary Matching Algorithm;BMA)在候選區塊大小越大時,其量測的失誤越嚴重,因此,以傳統BMA量測而得到較小失真值的區塊,其最小平方誤差(Mean-Square Error,MSE)比失真值較大的區塊之MSE值大的機率非常高。為了改善此一缺點,本論文進一步提出一個預測式的邊界相似度演算法,以增加度量的正確性,進而提昇影像品質。結果顯示利用預測式BMA可以比傳統的BMA提昇約0.5 dB之影像品質。 結合上述兩個方法進行實驗,其結果指出利用提出之方法可以比BMVR[10]方法之PSNR提昇約0.6 dB。
Noise combating has played an important role in data transmission for a long time. To reduce the resulted error of noise, many encoding mechanisms are included to a coding system. The flexible macroblock order (FMO) of H.264 is one of such mechanism, which intends to enhance the power of error correction by using the information around the corrupted block. In this thesis, we take advantage of FMO to propose a direction-oriented error concealment technique. The proposed method uses the motion information of the collocated block in the previous frame to select a more proper candidate block, so that a better error concealment can be obtained. To select a better block, boundary matching algorithm (BMA) is generally used. The traditional boundary matching algorithm, however, may lose its precision as the block size is increased. To raise the measured precision and to improve the visual quality of error concealment, we also propose a predicted boundary matching algorithm herein. Experimental results show that predicted BMA can raise about 0.5 dB of peak-signal to noise ratio (PSNR) value. By using the direction-oriented error concealment and predicted BMA, we can totally increase about0.6 dB compared with the BMVR[10].

Description

Keywords

H.264, 錯誤隱藏, 邊界相似度演算法, H.264, Error Concealment, Boundary Matching Algorithm (BMA)

Citation

Collections