一种复杂度可分级的帧内预测方法

Complexity-Scalable Intra-Frame Prediction Method

提出了一种复杂度可分级的帧内预测方法。利用相邻图像块纹理的相关性,选择概率大的部分预测模式对当前图像块进行处理,实现帧内预测复杂度的初步分级;然后根据复杂度冗余的大小,选择更多的预测模式进行处理,实现复杂度的精细分级。实验表明:本文方法不需要额外开销,可实现帧内预测复杂度的任意分级。

In this paper, a complexity-scalable intra-frame prediction method is proposed. Based on the texture correlation of the neighboring image-blocks, part of intra-frame prediction modes with high probability are selected for current image- block, which implement the primary scalability of computational complexity. And then, according to the redundancy of complexity, more prediction modes are chosen to be processed, which achieve refined scalability of complexity. Experiment results show that the proposed method needn’t extra payload, and can realize arbitrary complexity scalability of intra-frame prediction with high precision.