分形编码时间的线性控制和预测

Linear control and prediction of fractal coding time

以图像的布朗维数为纹理特征对编码中的图像块进行聚类和排序,实现了对每个值域块所需比较定义域块数目的精确控制。进一步,通过排除平坦块、使用均值图像建立定义域块库等技巧,得到了一种可以通过定义域块比较数目对编码时间进行线性控制和预测的快速分形编码方法。实验表明,与已有分类和聚类方法相比,在相同的压缩比下,本文方法可取得更好的加速效果和解码质量。

Taking the Brownian dimension as their texture feature, image blocks were clustered and sorted, to control the number of domain blocks to be compared with each range block in coding. Furthermore, by combining techniques such as excluding plain blocks and building domain pool from averaged image, a fast algorithm for fractal coding was obtained, in which the coding time can be linearly controlled and predicted by the number of domain blocks compared with each range block. Experiment results indicate that with the same compression ratio, the proposed method improves the speed of coding with even better quality of decoded image, in contrast to existing classification and clustering methods.