Sub-block size on impact of fundus image noise estimate

Image noise variance estimation is an important indicator to determine the image quality. In this paper. LMLSD algorithm based on Gaussian wave extraction is used for fundus image noise estimation, and it analyses the sub-block size effects for the performance of this algorithms. In the expriment, 6 different sub-block size are taken for the simulation and analysis of noise variance estimation respectively in 10 fundus images. Exprimental results show the accuracy of noise variance estimation which takes 5x5 sub-block size is better than others in 256*256 fundus images.