Abstract:

According to traditional blind source separation algorithm without taking into account noise, a new algorithm for blind separation of noisy image is proposed based on adaptive morphology in the curvelet transform domain. Curvelet transform has good performance in sparseness. Noisy image can be analyzed with curvelet transform, and denoised with adaptive denoising algorithm based on mathematical morphology. The separation matrix can be estimated by selecting the sparsest subband image. The mixed image can be separated thoroughly. Simulation results show that the proposed algorithm can achieve a better performance for blind source separation of noisy images.

Key words: blind source separation; sparse representation; curvelet transform; mathematical morphology; adaptive