블록 기반 영상에서 인근 정보의 유사성을 이용한 손실 블록 복구 기법
Restoration of lost blocks using similarity of adjacent information in block-based image
블록기반영상 BNM 손실블록 컴퓨터과학;
- 원문 URL
In this paper, we propose an algorithm for detection and correction of block error when block coded images are transmitted over error prone network. The images are transmitted over the wireless channel block by block. Due to differential coding of DC coefficient in JPEG images, losing a block, it may lose the following blocks. E. Chang reported that average packet loss rate in a wireless environment is 3.6% and the packet loss occurs in a bursty fashion. The proposed algorithm can use the information of not only adjacent regions, but also remote regions in images. Namely, instead of using retransmission scheme, the algorithm aims to reconstruct the lost information using correlation between the lost block and its neighbors. The algorithm executes following steps: searching, matching and filling. If it can't find a suitable candidate block, it applies blocking artifacts reduction scheme to the image block with minimized border error. It may be also used to reduce blocking artifacts in low bit rate JPEG images. First, the scheme finds a lost block and records a flag value in lost block marking table. And, it estimates DC coefficients of a lost block for decoding the following other blocks. For reconstruction of lost block, it finds proper candidate blocks in searching range and selects a matching block being minimized the squared error value of total border among them. It reconstructs the lost blocks with three steps: 1) Reconstructing the isolated lost blocks, 2) Reconstructing the diagonal lost blocks, 3) Reconstructing the other blocks. If isolated lost block of step 1 appears in step 2, it also reconstructs the lost block. Finally, it applies the blocking artifacts reduction scheme to improper matching blocks. We experiment its performance with different images for different loss rates. And, we magnify various sample regions for subjective evaluation and PSNR(Peak Signal-to-Noise Ratio) is used for objective evaluation. As compared with different algorithms, the algorithm can be executed very well with combining the proposed blocking artifacts reduction scheme. As a result, we obtained good results by experiments both objectively and subjectively. In the future, we will experiment with natural color JPEG images using the algorithm. And, we will research for error concealment algorithm of JPEG2000 and image transport protocol.