2002 Fiscal Year Final Research Report Summary
A Research on Fractal Image Coding for Mobile Image Communications
| Project/Area Number |
13450154
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (B)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Research Field |
情報通信工学
|
|---|
| Research Institution | Nagoya University |
|---|
Principal Investigator |
TANIMOTO Masayuki Nagoya University, Graduate School of Engineering, Professor, 工学研究科, 教授 (30109293)
|
|---|
| Project Period (FY) |
2001 – 2002
|
| Keywords | Fractal / Correlation / Collage theorem / Error resilience / Picture coding / Mobile communication |
|---|
| Research Abstract |
Through this research, the following results were obtained :
1. In fractal image decoding process, we investigated theoretically the mechanism of the propagation of image degradation caused by errors of received compressed data. We found that the propagated errors are about 2.4 times as the errors of compressed data regardless of the search area in encoding process.
2. We proposed a new coding method for error resilience by sending the image twice with the same amount of data as the single image. This is realized by using double-sized range blocks and a new criterion to determine the domain block.
3. In the proposed method, we determined the optimal shape of enlarged range blocks from the mean square error between range blocks and domain blocks. We confirmed the results from the viewpoint of the coding efficiency.
4. We also investigated the mechanism of the error detection in the decoding process. We found that the error detection using threshold sometimes fails, and this hampers error recovering and causes image degradation.
5. To prevent misdetection of error as 4, we introduced a pre-processing in the encoder, which selects only domain blocks that causes no error misdetection.
6. Using encoder and decoder based on 5, we succeeded to suppress the misdetection of errors.
We confirmed that the proposed method outperforms the conventional method with 2-3 dB gain of PSNR in an erroneous channel environment.
|
