Nafise Barzigar

SCoBeP: Dense image registration using sparse coding and belief propagation

Image registration as a basic task in image processing was studied widely in the literature. It is an important preprocessing step in different applications such as medical imaging, super resolution, and remote sensing. In this paper we proposed a novel dense registration method based on sparse coding and belief propagation. We used image blocks as features, and then we employed sparse coding to find a set of candidate points. To select optimum matches, belief propagation was subsequently applied on these candidate points. Experimental results show that the proposed approach is able to robustly register scenes and is competitive as compared to optical flow.

Decomposition Approach for Low-Rank Matrix Completion and Its Applications

In this paper, we describe a low-rank matrix completion method based on matrix decomposition. An incomplete matrix is decomposed into sub-matrices which are filled with a proposed trimming step and then are recombined to form a low-rank completed matrix. The divide-and-conquer approach can significantly reduce computation complexity and storage requirement. Moreover, the proposed decomposition method can be naturally incorporated into any existing matrix completion methods to attain further gain. Unlike most existing approaches, the proposed method is not based on norm minimization nor on SVD decomposition. This makes it possible to be applied beyond real domain and can be used in arbitrary fields, including finite fields. The effectiveness of our proposed method is demonstrated through extensive numerical results on randomly generated and real matrix completion problems and a concrete application-video denoising. The numerical experiments show that the algorithm can reliably solve a wide range of problems at a speed significantly faster than recent algorithms. In the proposed denoising approach, we present a patch-based video denoising algorithm by grouping similar patches and then formulating the problem of removing noise using a decomposition approach for low-rank matrix completion. Experiments show that the proposed approach robustly removes mixed noise such as impulsive noise, Poisson noise, and Gaussian noise from any natural noisy video. Moreover, our approach outperforms state-of-the-art denoising techniques such as VBM3D and 3DWTF in terms of both time and quality. Our technique also achieves significant improvement over time against other matrix completion methods.

Dense image registration using sparse coding and belief propagation

Image registration as a basic task in image processing was studied widely in the literature. It is an important preprocessing step in different applications such as medical imaging, super resolution, and remote sensing. In this paper we proposed a novel dense registration method based on sparse coding and belief propagation. We used image blocks as features, and then we employed sparse coding to find a set of candidate points. To select optimum matches, belief propagation was subsequently applied on these candidate points. Experimental results show that the proposed approach is able to robustly register scenes and is competitive as compared to optical flow.

Medical Image registration using sparse coding and belief propagation

Recently, various medical imaging such as CT and MRI imaging has been used more and more widely in clinical and medical research. As a result, there is an increasing interest in accurately relating information in different images for diagnosis, treatment, and the sake of basic science. As images are typically acquired at different times and often by different modalities, registering (or aligning) one image with another is not a simple task in general and it success will affect the effectiveness and accuracy of all subsequent analysis. We propose an efficient medical image registration method based on sparse coding and belief propagation for medical CT imaging. We used 3-D image blocks as features, and then we employed sparse coding to find a set of candidate voxels. To select optimum matches, belief propagation was subsequently applied on these candidate voxels. The outcome of belief propagation was interpreted as probabilistic map between candidate voxels and source voxel. We compared with the state-of-the-art of medical image registration, MIRT [1] and GP-Registration algorithm [2]. Our objective results based on RMSE (Root Mean Square Error) are smaller than those from MIRT and GP-Registration. Our results also proved the effectiveness of our algorithm in registering reference image to source image.

An efficient video denoising method using decomposition approach for low-rank matrix completion

Denoising as one of the most significant task in video processing was studied widely in the literature. We propose an efficient video denoising method based on decomposition approach for matrix completion. A noisy video is processed in blockwise manner and for each processed block we find similar blocks in other frames. The similar blocks then will stack together and unreliable pixels will remove using fast matrix completion method [1]. We demonstrate the effectiveness of our algorithm in removing the mixed noise through the results. Our results also proved the effectiveness of our algorithm in removing noise from regular structures. We also compare with other denoising technique using matrix completion. Our method results in comparable performance with significantly lower computation complexity.

A Video Super-Resolution Framework Using SCoBeP

Super-resolution as an exciting application in image processing was studied widely in the literature. This paper presents new approaches to video super-resolution based on sparse coding and belief propagation. First, find candidate match pixels on multiple frames using sparse coding and belief propagation. Second, incorporate information from these candidate pixels with weights computed using the nonlocal-means method in the first approach or using the sparse coding and belief propagation method in the second approach. The effectiveness of the proposed methods is demonstrated for both synthetic and real video sequences in the experiment section. In addition, the experimental results show that our models are naturally robust in handling super-resolution on video sequences affected by scene motions and/or small camera motions.

Real-time fMRI processing with physiological noise correction - Comparison with off-line analysis.

BACKGROUND While applications of real-time functional magnetic resonance imaging (rtfMRI) are growing rapidly, there are still limitations in real-time data processing compared to off-line analysis. NEW METHODS We developed a proof-of-concept real-time fMRI processing (rtfMRIp) system utilizing a personal computer (PC) with a dedicated graphic processing unit (GPU) to demonstrate that it is now possible to perform intensive whole-brain fMRI data processing in real-time. The rtfMRIp performs slice-timing correction, motion correction, spatial smoothing, signal scaling, and general linear model (GLM) analysis with multiple noise regressors including physiological noise modeled with cardiac (RETROICOR) and respiration volume per time (RVT). RESULTS The whole-brain data analysis with more than 100,000voxels and more than 250volumes is completed in less than 300ms, much faster than the time required to acquire the fMRI volume. Real-time processing implementation cannot be identical to off-line analysis when time-course information is used, such as in slice-timing correction, signal scaling, and GLM. We verified that reduced slice-timing correction for real-time analysis had comparable output with off-line analysis. The real-time GLM analysis, however, showed over-fitting when the number of sampled volumes was small. COMPARISON WITH EXISTING METHODS Our system implemented real-time RETROICOR and RVT physiological noise corrections for the first time and it is capable of processing these steps on all available data at a given time, without need for recursive algorithms. CONCLUSIONS Comprehensive data processing in rtfMRI is possible with a PC, while the number of samples should be considered in real-time GLM.

A robust super resolution method for video

Super resolution reconstruction produces a higher resolution image based on a set of low resolution images, taken from the same scene. Recently, many papers have been published, proposing a variety algorithms of video super resolution. This paper presents a new approach to video super resolution, based on sparse coding and belief propagation. First, find the candidate pixels on multiple frames using sparse coding and belief propagation. Second, exploit the similarities of candidate pixels using the Non-local Means method to average out the noise among similar patches. The experimental results show the effectiveness of our method and demonstrate its robustness to other super resolution methods.

Mixture noise removal in ultrasound images using SCoBeP and low-rank matrix completion

Denoising as one of the most significant tools in ultrasound imaging was studied widely in the literature. However, most existing ultrasound image denoising algorithms have assumed the additive white Gaussian noise. In this work, we propose two efficient ultrasound image denoising methods that can handle a noise mixture of various types. Our methods are based on SCoBeP [1] and low-rank matrix completion as follows. In our first method, a noisy image is processed in blockwise manner and for each processed block we find candidate match pixels on other images using sparse coding and belief propagation, where in our second algorithm, we use overlapped patches to further lower the computation complexity. The blocks centered around these candidate pixels then will stack together and unreliable pixels will be removed using fast matrix completion method [2]. We demonstrate the effectiveness of our algorithm in removing the mixed noise through the results. We also compare with other denoising technique using matrix completion. Our methods results in comparable performance with significantly lower computation complexity.

Nucleotide sequence alignment using sparse coding and belief propagation

Advances in DNA information extraction techniques have led to huge sequenced genomes from organisms spanning the tree of life. This increasing amount of genomic information requires tools for comparison of the nucleotide sequences. In this paper, we propose a novel nucleotide sequence alignment method based on sparse coding and belief propagation to compare the similarity of the nucleotide sequences. We used the neighbors of each nucleotide as features, and then we employed sparse coding to find a set of candidate nucleotides. To select optimum matches, belief propagation was subsequently applied to these candidate nucleotides. Experimental results show that the proposed approach is able to robustly align nucleotide sequences and is competitive to SOAPaligner [1] and BWA [2].