Haisen Li

Generative Bayesian Image Super Resolution With Natural Image Prior

We propose a new single image super resolution (SR) algorithm via Bayesian modeling with a natural image prior modeled by a high-order Markov random field (MRF). SR is one of the long-standing and active topics in image processing community. It is of great use in many practical applications, such as astronomical observation, medical imaging, and the adaptation of low-resolution contents onto high-resolution displays. One category of the conventional approaches for image SR is formulating the problem with Bayesian modeling techniques and then obtaining its maximum-a-posteriori solution, which actually boils down to a regularized regression task. Although straightforward, this approach cannot exploit the full potential offered by the probabilistic modeling, as only the posterior mode is sought. On the other hand, current Bayesian SR approaches using the posterior mean estimation typically use very simple prior models for natural images to ensure the computational tractability. In this paper, we present a Bayesian image SR approach with a flexible high-order MRF model as the prior for natural images. The minimum mean square error (MMSE) criteria are used for estimating the HR image. A Markov chain Monte Carlo-based sampling algorithm is presented for obtaining the MMSE solution. The proposed method cannot only enjoy the benefits offered by the flexible prior, but also has the advantage of making use of the probabilistic modeling to perform a posterior mean estimation, thus is less sensitive to the local minima problem as the MAP solution. Experimental results indicate that the proposed method can generate competitive or better results than state-of-the-art SR algorithms.

High dynamic range imaging by sparse representation

Abstract High dynamic range (HDR) imaging technology is becoming increasingly popular in various applications. A common approach to get an HDR image is the multiple exposed images fusion. However, the phenomenon of ghosting artifacts is brought in for the scene with non-static objects. This paper proposes a ghost-free HDR image synthesis algorithm that utilizes a sparse representation framework. Based on the dependency among adjacent low dynamic range (LDR) images and the sparsity of the moving object that leads to the ghost artifacts, we formulate the problem into two steps: moving object detection and ghost free HDR generation. In the moving object detection step, we formulate the problem as sparse representation due to the sparsity and instantaneous of the moving objects. In the HDR generation step, joint weighting is proposed to generate a ghost-free HDR image from the reference image. Experiments show that the proposed algorithm outperforms the state-of-the-art methods favorably on the textures and colors.

Joint Motion Deblurring with Blurred/Noisy Image Pair

Motion blurred images are widely existing when using a hand-held camera especially under the dim lighting conditions. Since edge information contained in the noisy image may be blurred by the motion blur, a blurred/noisy image pair captured under different exposure time can help to restore a sharp image. In the traditional deblurring methods based on blurred/noisy image pair, the deblurring process is in series with the denoising process, so that restoration result is sensitive to the denoised result. In this paper, we propose a robust algorithm to obtain the sharp image by fusing the blurred image and noisy image. By joint modeling the deblurring model and denoising model, the restoration result can be optimized via estimating the sharp image and blur kernel alternately in the proposed methods, and it is not sensitive to the denoised result benefited by the joint model. Experimental results demonstrated that the proposed method can achieve better performance compared with the state-of-the-art single image denoising methods, single image deblurring methods and blurred/noisy pair deblurring methods.

ARSAC: Efficient model estimation via adaptively ranked sample consensus

Abstract RANSAC is a popular robust model estimation algorithm in various computer vision applications. However, the speed of RANSAC declines dramatically as the inlier rate of the measurements decreases. In this paper, a novel Adaptively Ranked Sample Consensus(ARSAC) algorithm is presented to boost the speed and robustness of RANSAC. The algorithm adopts non-uniform sampling based on the ranked measurements to speed up the sampling process. Instead of a fixed measurement ranking, we design an adaptive scheme which updates the ranking of the measurements, to incorporate high quality measurements into sample at high priority. At the same time, a geometric constraint is proposed during sampling process to select measurements with scattered distribution in images, which could alleviate degenerate cases in epipolar geometry estimation. Experiments on both synthetic and real-world data demonstrate the superiority in efficiency and robustness of the proposed algorithm compared to the state-of-the-art methods.

Blind image deblurring by promoting group sparsity

Abstract Blind image deblurring aims to recover the sharp image from a blurred observation, which is an ill-posed inverse problem. Proper image priors for the unknown variables (i.e. latent sharp image and blur kernel) are crucial. Abundant previous methods have shown the effectiveness of the sparsity-based priors on both image gradients and the blur kernel. The correlation among the elements of the sparse variables is paid less attention, however. In this paper, we propose to handle the blind image deblurring problem by promoting group sparsity. The proposed group sparsity priors are based on the fact that the nonzero elements of natural image gradients and blur kernels tend to cluster in structured group pattern. Based on the proposed priors, we introduce proper algorithms to iteratively update latent image gradients and blur kernel, respectively. The proposed algorithms preserve the salient structures and smooth the minor components in image gradients and restrict the blur kernel in a domain of dynamic group sparse vector. To illustrate the reliability of the proposed algorithm, we conduct experiments to analyze the properties of the regularizers and the convergence property of the proposed algorithm. Experiments with both quantitative and visual comparison further prove the effectiveness of the proposed method.

Multi-image Deblurring Using Complement

The purpose of image restoration is to recover the latent image from the observed blurred images. In multi-image restoration, the input image’s quality directly affects the final deblurring result. In this paper, a blurry images restoration method based on the complement is proposed for the selection of images in multi-image restoration. First, we give a description of complement between different images. Then, the blurred images of maximum complement are selected by image selection method which is based on the iterative maximum complement. Finally, we use an existing deblurring method to estimate the latent image. We compare the new method with traditional methods, the experimental results clearly demonstrate the efficacy of the proposed method.

Blind Multi-frame Super Resolution with Non-identical Blur

Real world video super resolution is an challenging problem due to the complex motion field and unknown blur kernel. Although multi-frame super resolution has been extensively studied in past decades, it still remained problems and always assumed that the blur kernels were identical in different frames. In this paper, we propose an novel blind multi-frame super resolution method with non-identical blur. To estimate blur kernels of different frames, we propose using salient edges selection method for more accurate kernel estimation. The whole process of estimation is based on Hyper-Laplacian prior, and iterative value updating through a multi-scale process. After the kernels of different frames are estimated, the high resolution frame is reconstructed using a cost function. The proposed method can obtain superior results, and outperforms the state of the art in the experiments through subjective and objective evaluation.

A Dim Small Target Detection Method Based on Spatial-Frequency Domain Features Space

The target detection, especially extracting low SNR potential targets and stars from the star images, plays as a key technology in the space debris surveillance. Due to the complexity of the imaging environment, the detection of dim small targets in star images faces many difficulties, including low SNR and rare unstable features. This paper proposes a dim small target detection method based on the high dimensional spatial-frequency domain features extracted by filter bank, and training the support vector machine (SVM) classifier. The experimental results demonstrate that the proposed method exceeds the state-of-the-art on the ability to detect low SNR targets.

Statistical learning modeling method for space debris photometric measurement

Photometric measurement is an important way to identify the space debris, but the present methods of photometric measurement have many constraints on star image and need complex image processing. Aiming at the problems, a statistical learning modeling method for space debris photometric measurement is proposed based on the global consistency of the star image, and the statistical information of star images is used to eliminate the measurement noises. First, the known stars on the star image are divided into training stars and testing stars. Then, the training stars are selected as the least squares fitting parameters to construct the photometric measurement model, and the testing stars are used to calculate the measurement accuracy of the photometric measurement model. Experimental results show that, the accuracy of the proposed photometric measurement model is about 0.1 magnitudes.

ARSAC: Robust Model Estimation via Adaptively Ranked Sample Consensus

RANSAC is a popular model estimation algorithm in various of computer vision applications. However, it easily gets slow as the inlier rate of the measurements declines. In this paper, a novel Adaptively Ranked Sample Consensus (ARSAC) algorithm is presented to boost the speed and robustness of RANSAC. Our algorithm adopts non-uniform sampling based on the ranked measurements. We propose an adaptive scheme which updates the ranking of the measurements on each trial, to incorporate high quality measurement into sample at high priority. We also design a geometric constraint during sampling process, which could alleviate degenerate cases caused by non-uniform sampling in epipolar geometry. Experiments on real-world data demonstrate the effectiveness and robustness of the proposed method compared to the state-of-the-art methods.