1. Image denoising based on adaptive sparse representation
- Author
-
Yang Min Li, Jinwu Xu Jianhong, and Guodong Wang
- Subjects
K-SVD ,business.industry ,Noise reduction ,ComputingMethodologies_IMAGEPROCESSINGANDCOMPUTERVISION ,Basis pursuit ,Pattern recognition ,Sparse approximation ,Non-local means ,Matching pursuit ,symbols.namesake ,Additive white Gaussian noise ,Gaussian noise ,Computer Science::Computer Vision and Pattern Recognition ,symbols ,Computer vision ,Artificial intelligence ,business ,Mathematics - Abstract
Image denoising usually needs to estimate noise variance. In order to avoid estimating the noise variance and remove the white Gaussian noise from image, a denoising method based on adaptive sparse representation was proposed. It trains the initialized dictionary based on training samples constructed from noised image. The training process is finished by an iteration algorithm which alternates between adaptive sparse representation and dictionary update. Based on the trained dictionary, noise reduction is conducted through adaptive sparse representation of the noised image. Compared with adaptive Wiener filtering and adaptive denoising based on Basis Pursuit, the proposed method could remain more image details and have better performance. With the proposed method, laser electronic speckle interference image could be enhanced and its interference fringe became clearer.
- Published
- 2010