Adaptive Directional Lifting Transform based Optimization to Extreme Visual Recovery

Rank minimization is giving the solutions to the problems in image denoising, machine learning, computer vision and compressed sensing applications. The corrupted image will not convey the original information and during transmission the information may further tampered. The image is corrupted with small amount of noise, it may be recovered with the conventional methods. The amount of noise in the image is more, the existing methods are unable to recover the corrupted image. Here an Adaptive Directional Lifting Wavelet Transform (ADLWT) is used along with the familiar Alternating Direction Method of Multipliers optimization method. The proposed method is able to recover the corrupted image that contains high percentage of missing observations. The objective of the work is to provide an efficient method to recover the damaged pixels in the image and to provide the optimal λ value which can recover with high accuracy and efficiency with respect to quality metrics. The proposed method is evaluated with the familiar assessment measures like Mean Square Error (MSE), Peak Signal to Noise Ratio (PSNR) and Structural Similarity (SSIM). The recorded values are better when compared with the existing methods.