摘要
针对图像稀疏分解中原子生成速度慢的难题,本文提出了原子生成的一种快速算法。首先根据原子的尺度把原子分成两大类,一类是小原子,一类是大原子。在此基础上,对于小原子,由于其能量集中分布在一个小的范围,所以利用一个小范围生成的局部原子代替整个原子。对于大原子,先生成一个相对应的较小原子,然后通过插值方法生成大原子。实验结果表明,在重建图像的质量没有任何改变的条件下,当图像大小为256×256时,提出的算法使原子生成的速度提高了20多倍。
It's one of main problems in image sparse decomposition that the atom construction process is very time -consuming. To overcome this key problem a new fast algorithm is presented. At first all atoms are divided into two kinds, small atoms and big atoms. For small atoms, the construction of whole atom can be substituted by the construction of a small part of atom because of the concentration of atom energy in a limited region. On the other hand, the construction of big atom can be done by interpolating small atom which is constructed according to the parameters of the big atom except with a smaller scale. Experimental results show that, when the size of the images is 256 × 256, the proposed algorithm speeds up a little more than 20 times the atom construction process without any loss of the reconstructed image quality compared with the original method.
出处
《电讯技术》
2005年第6期12-16,共5页
Telecommunication Engineering
基金
国家留学基金资助项目(21851039)
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
稀疏分解
过完备原子库
快速算法
Image processing
Sparse decomposition
Over - complete dictionary of atoms
Fast algorithm