摘要
针对图像稀疏分解的计算时间复杂度非常高这个问题,提出了分块自适应图像稀疏分解算法。该算法根据稀疏分解计算时间复杂度和待分解图像大小之间的关系,把待分解图像分成互不重叠的小块,然后对每个小块图像进行稀疏分解。根据每一块的复杂程度,自适应地决定稀疏分解的结束。实验结果表明,在分解原子个数相近或相同的条件下,新算法对稀疏分解后重建图像比在整幅图像上进行稀疏分解重建的图像质量下降0.5 dB,但计算速度提高了约15倍。
The computational burden in image sparse decomposition process is very huge. To deal with this problem, an adaptive block -based sparse decomposition algorithm is propesed. Based on the relation between computational burden and the image size, the new algorithm divides the whole image into small blocks which are not superpesed, then sparse decomposition of one image is transformed into sparse decomposition of small blocks of the original image. Experimental results show that with approximato number of atoms, the PSNR value of the image constructed by the new algorithm is degraded by about 0.5 dB, but the computing speed is improved by about 15 times, compared with the original whole image sparse decomposition method.
出处
《电讯技术》
2006年第4期63-67,共5页
Telecommunication Engineering
基金
四川省重点科技计划项目(04GG021-020-5
03GG006-005-2)
教育部留学回国人员科研启动基金资助项目(教外司[2004]527号)
关键词
图像处理
稀疏表示
稀疏分解
匹配追踪
image processing
sparse representation
sparse decomposition
matching pursuit （MP）
