Phase Retrieval Based RIDNet Denoising Algorithm for Color Image

The phase retrieval problem refers to recovering the original signal from only amplitude measurements. The detector CCD can only seek the intensity information in the diffraction field, and the phase is generally lost. The phase retrieval problem is non-linear, and it is difficult to reconstruct the original from only the amplitude of the color image. The phase retrieval algorithm, based on RIDNet denoising(RIDNetPR) is proposed to deal with color image reconstruction. First, the features are extracted using RIDNet, and multiple Enhancement Attention Modules (EAM) are repeatedly stacked for color image denoising tasks to ensure the ability of the deep learning network for feature extraction. Regularization terms constrain the network by improving the recovery accuracy. Second, the Alternating Direction Method of Multipliers (ADMM) is utilized to reconstruct the original color images. Compared with traditional phase retrieval algorithms, the experimental results show that the proposed algorithm improves the Peak Signal-to-Noise Ratio (PSNR) and Structure Similarity Index Measure(SSIM) values at different Gaussian noise levels. The effectiveness of the proposed algorithm is illustrated for color image.