A new algorithm is proposed for compressed sensingmagnetic resonance imaging (CS-MRI). The l <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">p</sub> -norm (0 <; p ≤ 1) based adaptive regularization model used MRI. established by using a novel iterative shrinkage scheme. In the iteration, quasi-Newton method employed. shrinkage, threshold defined varyingly. Also, parameter selected dynamically in algorithm. Comparing with some certain...

Multi-scale architectures and attention modules have shown effectiveness in many deep learning-based image de-raining methods. However, manually designing integrating these two components into a neural network requires bulk of labor extensive expertise. In this article, high-performance multi-scale attentive architecture search (MANAS) framework is technically developed for de-raining. The proposed method formulates new space with multiple flexible that are favorite to the task. Under space,...

Single image de-raining is an important and highly challenging problem. To address this problem, some depth or density guided single-image methods have been developed with encouraging performance. However, these individually use the to guide network conduct de-raining. In paper, a novel <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">joint de-raining</i> (JDDGD) method technically developed. The JDDGD starts...

Conventional face anti-spoofing methods might be poorly generalized to unseen data distributions. Thus, we improve the generalization of spoof detection from multi-domain feature disentanglement. Specially, a two-branch convolutional network is proposed separate spoof-specific features and domain-specific images explicitly. The are further used for live vs. classification. To minimize correlation among these two features, present cross-adversarial training scheme, which requires each branch...

Most of the face anti-spoofing methods improve generalization capability by adversarial domain adaptation via training source and target data jointly. However, considering privacy, it is impractical in application. Hence, we propose a data-free adaptative framework to optimize network without using labeled modeling into problem learning with noisy labels. To obtain more reliable pseudo labels, dynamic images background capture motion divergences between real attack faces. Nonetheless,...

In this letter, robust sparse signal recovery is considered in the presence of symmetric α-stable distributed noise. An M-estimate type model constructed by approximating location score function A reweighed iterative hard thresholding algorithm proposed to recover signal. The basis functions for approximation and performance are discussed. Simulations given demonstrate validity our results.

Rain removal is a highly demanding task because rainy image in computer lacks discriminative information to distinguish the details from rain streaks. In this paper, we present new High-Low-Frequency Guided De-raining (HLFGD) method remove streaks clearly while reserve details. Specifically, proposed HLFGD built with three network branches, namely global-structure branch, de-raining and edge-detail which achieve collaboration by concatenating intermediate features. Among them, branches aim...

Abstract Stereo matching, which is a key problem in computer vision, faces the challenge of radiometric distortions. Most existing stereo matching methods are based on simple cost algorithms and appear mismatch under It necessary to improve robustness accuracy algorithms. A novel encoding pattern proposed for matching. In pattern, each windows grey image gradient images divided into several isoline‐like sets with different radii. Then, pixel pairs defined sets. An function used decide...

In this paper, a segmentation-based mutual information image registration algorithm is proposed. Because of noise-like speckle, calculated on segmented results instead original images. Instead using pixel-based MRF segmentation, region-based RAG-MRF applied oversegmentation formed by Watershed transformation the gradient SAR The extrema indicate desired parameters. conception introduced followed principle segmentation and watershed transformation. Experiment showed it an efficient approach for image.

3D Morphable Model (3DMM) is a statistical tool widely employed in reconstructing face shape. Existing methods are aimed at predicting 3DMM shape parameters with single encoder but suffer from unclear distinction of different attributes. To address this problem, Two-Pathway Encoder-Decoder Network (2PEDN) proposed to regress the identity and expression components via global local pathways. Specifically, each 2D image cropped into details as inputs for corresponding 2PEDN trained predict two...

As a prior knowledge, non‐local self‐similarity (NSS) has been widely utilised in ill‐posed problems. Actually, similar textures appear not only single scale, but also different scales. Unlike most existing patch‐based methods that explore NSS the same multi‐scale patches based image denoising algorithm is proposed this study. The authors have designed strategy to expand search space of block‐matching, which will increase probability finding more patches. After that, weighted nuclear norm...

Denoising Magnetic Resonance (MR) image as a challenging work has been widely concerned. Minimax concave penalty (MCP) is an effective denoising method, which can effectively remove the additive Gaussian noise. However, existing low rank matrix approximation methods not Rician noise in magnetic resonance imaging (MRI). To solve this problem, MR method based on extended Differential of Gssian (DoG) filter and adaptive minimax-concave proposed. In new MCP model used to improve quality image....

Abstract In our paper, we introduce the sparse-friendly distillation framework as an effective training strategy for knowledge distillation. While model sparsity techniques have been widely adopted to reduce overhead, sparse student models often struggle achieve good performance in To address this issue, leverages observation that exhibit different behaviors foreground and background features. We separate these features using pooling apply mean squared error (MSE) feature Furthermore,...

Transmission of videos in error prone environments may lead to video corruption or loss. Therefore concealment at the decoder side has be applied. Commonly techniques make use surrounding correctly received image data motion information for concealment. In this paper, a novel spatio-temporal boundary matching algorithm (STBMA) by exploiting both spatial and temporal reconstruct lost vectors (MV) is proposed, also introduce new smoothness measurement. By using vector that found proposed...