Joint Denoising of Stereo Images Using 3D CNN
Malek Khammassi, Mounir Kaaniche, Amel Benazza‐Benyahia · 2021
In this paper, we propose a joint denoising algorithm to reduce noise in stereo images. The proposed algorithm is a modified version of a recent deep CNN-based single image denoising approach. Our contribution consists in exploiting both intra- and inter-view correlations thay exist between the left and right views. More precisely, we resort to a residual learning network to extract the noise and apply batch normalization to stabilize and speed up the training. Compared with state-of-the-art methods, our method shows promising results in terms of noise reduction in stereo images.