Tao Dai

Entropy-based bilateral filtering with a new range kernel

An entropy-based bilateral filter for image denoising.A new range distance derived from a clean image is robust to noise.Adaptive range parameter considers the local structures of images.Experiments demonstrate the proposed method significantly outperforms the standard bilateral filter. Bilateral filter (BF) is a well-known edge-preserving image smoothing technique, which has been widely used in image denoising. The major drawback of BF is that its range kernel is sensitive to noise. To address this issue, we propose an entropy-based BF (EBF) with a new range kernel which contains a new range distance. The new range distance is robust to noise by exploiting the information from the denoised estimate and the corresponding method noise, i.e., the difference between the noisy image and its denoised estimate. Moreover, in order to consider the local statistics of images, local entropy is applied to adaptively guide the range parameter selections. This allows our method to adapt to the images with different characteristics. Experimental results demonstrate that the proposed EBF significantly outperforms the standard BF in terms of both quantitative metrics and subjective visual quality.

A generic denoising framework via guided principal component analysis

Though existing state-of-the-art denoising algorithms, such as BM3D, LPG-PCA and DDF, obtain remarkable results, these methods are not good at preserving details at high noise levels, sometimes even introducing non-existent artifacts. To improve the performance of these denoising methods at high noise levels, a generic denoising framework is proposed in this paper, which is based on guided principle component analysis (GPCA). The propose framework can be split into two stages. First, we use statistic test to generate an initial denoised image through back projection, where the statistical test can detect the significantly relevant information between the denoised image and the corresponding residual image. Second, similar image patches are collected to form different patch groups, and local basis are learned from each patch group by principle component analysis. Experimental results on natural images, contaminated with Gaussian and non-Gaussian noise, verify the effectiveness of the proposed framework.

Referenceless quality metric of multiply-distorted images based on structural degradation

Abstract Multiply-distorted images, that is, distorted by different types of distortions simultaneously, are so common in real applications. This kind of images contain multiple overlaying stages (e.g., acquisition, compression and transmission stage). Each stage will introduce a certain type of distortion, for example, sensor noise in acquisition stage and compression artifacts in compression stage. However, most current blind/no-reference image quality assessment (NR-IQA) methods are specifically designed for singly-distorted images, thus resulting in their deficiency in handling multiply-distorted images. Motivated by the hypothesis that human visual system (HVS) is adapted to the structural information in images, we attempt to assess multiply-distorted images based on structural degradation. To this end, we use both first- and high-order image structures to design a novel referenceless quality metric for multiply-distorted images. Specifically, we leverage the quality-aware features extracted from both the gradient-magnitude map and contrast-normalized map, and further improve the performance by making use of redundancy of features with random subspace method. Experimental results on popular multiply-distorted image databases verify the outstanding performance of the proposed method.

Robust Survey Aggregation with Student-t Distribution and Sparse Representation

Most existing survey aggregation methods assume that the sample data follow Gaussian distribution. However, these methods are sensitive to outliers, due to the thin-tailed property of Gaussian distribution. To address this issue, we propose a robust survey aggregation method based on Student-t distribution and sparse representation. Specifically, we assume that the samples follow Student-t distribution, instead of the common Gaussian distribution. Due to the Student-t distribution, our method is robust to outliers, which can be explained from both Bayesian point of view and non-Bayesian point of view. In addition, inspired by James-Stain estimator (JS) and Compressive Averaging (CAvg), we propose to sparsely represent the global mean vector by an adaptive basis comprising both dataspecific basis and combined generic basis. Theoretically, we prove that JS and CAvg are special cases of our method. Extensive experiments demonstrate that our proposed method achieves significant improvement over the state-of-the-art methods on both synthetic and real datasets.

Student-t Process Regression with Student-t Likelihood

Gaussian Process Regression (GPR) is a powerful Bayesian method. However, the performance of GPR can be significantly degraded when the training data are contaminated by outliers, including target outliers and input outliers. Although there are some variants of GPR (e.g., GPR with Student-t likelihood (GPRT)) aiming to handle outliers, most of the variants focus on handling the target outliers while little effort has been done to deal with the input outliers. In contrast, in this work, we aim to handle both the target outliers and the input outliers at the same time. Specifically, we replace the Gaussian noise in GPR with independent Student-t noise to cope with the target outliers. Moreover, to enhance the robustness w.r.t. the input outliers, we use a Student-t Process prior instead of the common Gaussian Process prior, leading to Student-t Process Regression with Student-t Likelihood (TPRT). We theoretically show that TPRT is more robust to both input and target outliers than GPR and GPRT, and prove that both GPR and GPRT are special cases of TPRT. Various experiments demonstrate that TPRT outperforms GPR and its variants on both synthetic and real datasets.

Compressed Sensing Performance of Binary Matrices with Binary Column Correlations

This paper studies a class of binary matrices with correlations between distinct columnsequal to zero or one, which has reported comparable performance with random matrices inrecent studies of compressed sensing. For such matrix, we analyze its structure propertyand provide an improved performance estimation.