基于相似比的变邻域搜索的快速分形编码算法

为了解决基本分形图像编码算法中的编码过程特别耗时问题,通过定义每个range块和domain块的相似比,建立它与匹配均方根误差间的关系不等式,可把寻找range块的最佳匹配domain块的全局搜索变为近邻搜索.鉴于在自仿射变换下最优匹配块间的相似比值应该接近,但它们间的远近程度不一致,因此,每个range块的最优匹配块搜索范围应限制在与其相似比值接近的domain块变邻域内.四幅图像的仿真结果表明,它确实能够在PSNR降低0.103d B(其结构相似性SSIM值仅下降0.0004)的情况下,平均耗时仅为基本分形编码算法的38.97%左右,而且也优于可选特征算法,实现了加快编码过程速度的目标.

Variable neighbor search of fast fractal encoding algorithm based on similar ratio

In order to curtail the exhaustive encoding time of basic fractal image encoding algorithm. By defining the similar ratio of each range block and domain block, during the best matching search process, realizing the nearest neighbor search instead of the entire search. In view of the similar ratio of the optimal matching block should be adjacent based on the oneself affine transformation,but the distance between the range block and domain block is inconsistent, therefore, the optimal matching search scope for each range block block should be the variable neighbor of the domain block having the closest similar ratio to the input range block being encoded. Simulation results of four test images show that average time of the proposed scheme is only about 38.97% while there is averagely the PSNR decrease of 0.103dB(the structural similarity decrease of 0.0004), in comparison with the basic fractal algorithm.Moreover, it is better than the optional feature algorithm,the proposed algorithm achieve an objective of speed up the encoding process.