| 摘要 |
For coding with intra-prediction according to the H.264/AVC standard, reconstructed pixels from adjacent blocks are used to perform prediction of a target block, so that the prediction needs to be performed by referencing spatially-separate pixels. As such, the prediction accuracy is not always sufficient.
An intra-prediction technique is proposed, whereby for each pixel in the block an adjacent pixel is always used as reference. During coding, the difference value (prediction residue) between the adjacent pixels is calculated in a certain direction (predicting direction) and encoded with the predicting direction. During coding of a pixel, adjacent pixels may have not been encoded yet. In that case, the difference value between the original values of the adjacent pixels is encoded. On the other hand, in the decoder, processing to add a value produced by decoding the previous pixel value and the prediction residue is repeated in a predicting direction to obtain a decoded image. This can perform highly-efficient coding due to the strong correlation between adjacent pixels. |
