會議論文
| 學年 | 98 |
|---|---|
| 學期 | 1 |
| 發表日期 | 2009-12-03 |
| 作品名稱 | A Fast Intra Prediction Based on Haar Transform in H.264/AVC |
| 作品名稱（其他語言） | |
| 著者 | Yen, Shwu-Huey; Lin, meng-ju; Hwan, shin-jia; Li, tai-kuang |
| 作品所屬單位 | 淡江大學資訊工程學系 |
| 出版者 | Taipei : Institute of electrical and electronics engineers (IEEE) |
| 會議名稱 | JCPC 2009 Joint Conferences on Pervasive Computing |
| 會議地點 | Taipei, Taiwan |
| 摘要 | H.264/AVC is one of the state of the art in video compressing standards. It has advantages of low bit-rate and high image quality. However, the omplexity of H.264/AVC encoder limits applications in the real time video communication. In this study, a fast intra prediction with Haar transform algorithm (FIPHTA) is proposed to simplify the complexity of luma intra prediction. For each luma macroblock (MB), a Haar transform is first executed. By observing LL coefficients, this MB will be classified as 16-MB, I4-MB, or both. An I16-MB indicates that this MB is smooth and only Intra 16×16 (I16) prediction modes are to be considered. An I4-MB indicates that this MB is textured and only Intra 4×4 (I4) prediction modes are to be considered. And the rest of MBs (both) have to consider both prediction modes as in the standard H.264/AVC. Next, to choose an I16 prediction mode for a MB, the vertical and horizontal energies of the LL coefficients are calculated to reduce candidate modes. Similarly, the pixel-based vertical and horizontal energies are calculated for a 4×4 block to reduce I4 candidate modes. The simulation showed that the proposed algorithm can maintain the similar mode selection results comparing to the full search algorithm of H.264 but with a remarkable computation deduction. |
| 關鍵字 | H.264/AVC;Haar transform;Intra predictio |
| 語言 | en |
| 收錄於 | |
| 會議性質 | 國際 |
| 校內研討會地點 | 淡水校園 |
| 研討會時間 | 20091203~20091205 |
| 通訊作者 | |
| 國別 | |
| 公開徵稿 | Y |
| 出版型式 | |
| 出處 | Proceedings of the 2009 joint conferences on pervasive computing, pp.161-166 |
| 相關連結 |
機構典藏連結 ( http://tkuir.lib.tku.edu.tw:8080/dspace/handle/987654321/73167 ) |