TV+TV²Regularization with Nonconvex Sparseness-Inducing Penalty for Image Restoration

Abstract

In order to restore the high quality image, we propose a compound regularization method which combines a new higher-order extension of total variation (TV+TV2) and a nonconvex sparseness-inducing penalty. Considering the presence of varying directional features in images, we employ the shearlet transform to preserve the abundant geometrical information of the image. The nonconvex sparseness-inducing penalty approach increases robustness to noise and image nonsparsity. In what follows, we present the numerical solution of the proposed model by employing the split Bregman iteration and a novelp-shrinkage operator. And finally, we perform numerical experiments for image denoising, image deblurring, and image reconstructing from incomplete spectral samples. The experimental results demonstrate the efficiency of the proposed restoration method for preserving the structure details and the sharp edges of image.

Simultaneous Super-Resolution and Distortion Correction for Single-shot EPI DWI using Deep Learning

“…During the training of the main network, the FA calculation branch calculates FA maps using the super-resolution outputs and the high-resolution targets, respectively. The differences of the FA results are termed as the FA loss: For the loss function design, we combine the structural consistency index (SSIM) [48], pixel-wise difference (ℒ 1 ), image edge preservation (Gradient loss and gradient map loss) [49], the proposed FA loss and the cross-entropy GAN loss (ℒ GAN ). The loss function can be written as Here ∇ refers to gradient operator, and ℒ GAN refers to: …”

Quick Overview of Face Swap Deep Fakes

“…No additional enhancements were made to Faceswap, so the cost function proves that the result is based only on image reconstruction. The developers implemented such cost functions as L1 (mean absolute error), L2 (mean squared error), Logcosh, generalized loss [42], L-inf norm, DSSIM (difference of structural similarity), GMSDLoss (gradient magnitude similarity deviation loss) [43], and GradientLoss [44].…”

Compressed Sensing MRI Using Discrete Nonseparable Shearlet Transform and FISTA

“…The combination of two or more sparsifying transforms is conveniently implemented using analysis based reconstruction algorithms, such as the Manuscript Split Bregman Reconstruction Algorithm (SBRA) [6]- [8] and the Alternating Direction of Multipliers algorithms [9]. The use of multiple transforms leads to a number of regularization parameters and increases the algorithm complexity.…”