关键词: 图像序列超分辨率 充分帧数 差异度量

Abstract:

Sufficient Number of Low-resolution Images （SNL） estimation is a key problem in the application of super-resolution （SR） in image sequences. This paper presents a practical SNL estimation algorithm. The basic idea is to utilize the convergence of the reconstructed high-resolution images （HRI） coonsidering the increase of SNL. First,the sub-sequences with different numbers of frames are put into the SR procedure. Each sub-sequence corresponds to  a HRI, so we have a sequence of HRIs. Second, the differences between the HRIs are measured. Finally, SNL is estimated though the analyses of the diffe  rence curve of the HRIs. The experiments on real data demonstrate that the proposed algorithm estimates the SNL accurately and stably. The presented met hod can be used to support the practical application of the SR technology.

Key words: super-resolution in image sequences;sufficient number of low-resolution images;difference measurement