164782623-Matlab-2013-pdf

VENSOFT Technologies, www.ieeedeveloperslabs.in Email: [email protected] Contact: 9448847874


MATLAB PROJECT TITLES 2013-2014

317. Non locally Centralized Sparse Representation For Image Restoration

Abstract: The sparse representation models code an image patch as a linear combination of afew atoms chosen out from an over-complete dictionary, and they have shown promisingresults in various image restoration applications. However, due to the degradation of theobserved image (e.g., noisy, blurred and/or downsampled), the sparse representations byconventional models may not be accurate enough for a faithful reconstruction of the originalimage. To improve the performance of sparse representation based image restoration, in thispaper the concept of sparse coding noise is introduced, and the goal of image restoration turnsto how to suppress the sparse coding noise. To this end, we exploit the image nonlocal self-similarity to obtain good estimates of the sparse coding coefficients of the original image, andthen centralize the sparse coding coefficients of the observed image to those estimates. The so-called nonlocally centralized sparse representation (NCSR) model is as simple as the standardsparse representation model, while our extensive experiments on various types of imagerestoration problems, including denoising, deblurring and super-resolution, validate thegenerality and state-of-the-art performance of the proposed NCSR algorithm.

Keywords: Sparse representation, image restoration, nonlocal similarity

318. Sparse Representation Based Image Interpolation With Nonlocal AutoregressiveModeling

Abstract: Sparse representation has proven to be a promising approach to image super-resolution, where the low resolution (LR) image is usually modeled as the down-sampledversion of its high resolution (HR) counterpart after blurring. When the blurring kernel is theDirac delta function, i.e., the LR image is directly down-sampled from its HR counterpartwithout blurring, the super-resolution problem becomes an image interpolation problem. Insuch case, however, the conventional sparse representation models (SRM) become lesseffective because the data fidelity term will fail to constrain the image local structures. Innatural images, fortunately, the many nonlocal similar patches to a given patch could providenonlocal constraint to the local structure. In this paper we incorporate the image nonlocal self-similarity into SRM for image interpolation. More specifically, a nonlocal autoregressive model(NARM) is proposed and taken as the data fidelity term in SRM. We show that the NARMinduced sampling matrix is less coherent with the representation dictionary, and consequentlymakes SRM more effective for image interpolation. Our extensive experimental resultsdemonstrated that the proposed NARM based image interpolation method can effectivelyreconstruct the edge structures and suppress the jaggy/ringing artifacts, achieving the bestimage interpolation results so far in term of PSNR as well as perceptual quality metrics such asSSIM and FSIM.

Index Terms: Image interpolation, super-resolution, sparse representation, nonlocalautoregressive model.



319.Removing Atmospheric Turbulence Via Space-Invariant De convolution

Abstract: To correct geometric distortion and reduce space and time-varying blur, a newapproach is proposed in this paper capable of restoring a single high-quality image from a givenimage sequence distorted by atmospheric turbulence. This approach reduces the space andtime-varying deblurring problem to a shift invariant one. It first registers each frame tosuppress geometric deformation through B-spline-based nonrigid registration. Next, a temporalregression process is carried out to produce an image from the registered frames, which can beviewed as being convolved with a space invariant near-diffraction-limited blur. Finally, a blinddeconvolution algorithm is implemented to deblur the fused image, generating a final output.Experiments using real data illustrate that this approach can effectively alleviate blur anddistortions, recover details of the scene, and significantly improve visual quality.

Index Terms: Image restoration, atmospheric turbulence, nonrigid image registration, pointspread function, sharpness metric

320. SAIF-Ly Boost De noising Performance

Abstract: Spatial domain image filters (e.g. bilateral filter, NLM, LARK) have achieved greatsuccess in denoising. However, their overall performance has not generally surpassed theleading transform domain based filters (such as BM3D). One important reason is that spatialdomain filters lack an efficient way to adaptively fine tune their denoising strength; somethingthat is relatively easy to do in transform domain method with shrinkage operators. In the pixeldomain, the smoothing strength is usually controlled globally by, for example, tuning aregularization parameter. In this paper, we propose SAIF1 (Spatially Adaptive IterativeFiltering), a new strategy to control the denoising strength locally for any spatial domainmethod. This approach is capable of filtering local image content iteratively using the givenbase filter, while the type of iteration and the iteration number are automatically optimizedwith respect to estimated risk (i.e. mean-squared error). In exploiting the estimated local SNR,we also present a new risk estimator which is different than the often-employed SURE methodand exceeds its performance in many cases. Experiments illustrate that our strategy cansignificantly relax the base algorithm’s sensitivity to its tuning (smoothing) parameters, andeffectively boost the performance of several existing denoising filters to generate state-of-the-art results under both simulated and practical conditions.

Index Terms: Image denoising, spatial domain filter, risk estimator, SURE, pixel aggregation

321. Image Signature: Highlighting Sparse Salient Regions

Abstract: We introduce a simple image descriptor referred to as the image signature. We show,within the theoretical framework of sparse signal mixing, that this quantity spatiallyapproximates the foreground of an image. We experimentally investigate whether thisapproximate foreground overlaps with visually conspicuous image locations by developing a



saliency algorithm based on the image signature. This saliency algorithm predicts humanfixation points best among competitors on the Bruce and Tsotsos [1] benchmark data set anddoes so in much shorter running time. In a related experiment, we demonstrate with a changeblindness data set that the distance between images induced by the image signature is closer tohuman perceptual distance than can be achieved using other saliency algorithms, pixel-wise, orGIST [2] descriptor methods.

Index Terms: Saliency, visual attention, change blindness, sign function, sparse signal analysis.

322. Nonparametric Bayesian Dictionary Learning for Analysis of Noisy and IncompleteImages.

Abstract: Nonparametric Bayesian methods are considered for recovery of imagery based uponcompressive, incomplete, and/or noisy measurements. A truncated beta-Bernoulli process isemployed to infer an appropriate dictionary for the data under test and also for imagerecovery. In the context of compressive sensing, significant improvements in image recoveryare manifested using learned dictionaries, relative to using standard orthonormal imageexpansions. The compressive-measurement projections are also optimized for the learneddictionary. Additionally, we consider simpler (incomplete) measurements, defined bymeasuring a subset of image pixels, uniformly selected at random. Spatial interrelationshipswithin imagery are exploited through use of the Dirichlet and probit stick-breaking processes.Several example results are presented, with comparisons to other methods in the literature.

Index Terms: Bayesian nonparametrics, compressive sensing, dictionary learning, factoranalysis, image denoising, image interpolation, sparse coding.

323. Patch-Based Near-Optimal Image Denoising

Abstract: In this paper, we propose a denoising method motivated by our previous analysis ofthe performance bounds for image denoising. Insights from that study are used here to derivea high-performance practical denoising algorithm. We propose a patch-based Wiener filter thatexploits patch redundancy for image denoising. Our framework uses both geometrically andphotometrically similar patches to estimate the different filter parameters. We describe howthese parameters can be accurately estimated directly from the input noisy image. Ourdenoising approach, designed for near-optimal performance (in the mean-squared error sense),has a sound statistical foundation that is analyzed in detail. The performance of our approach isexperimentally verified on a variety of images and noise levels. The results presented heredemonstrate that our proposed method is on par or exceeding the current state of the art, bothvisually and quantitatively.

Index Terms: Denoising bounds, image clustering, image denoising, linear minimum mean-squared-error (LMMSE) estimator, Wiener filter.



324 .Accelerated Hypothesis Generation for Multistructure Data via Preference Analysis.

Abstract: Random hypothesis generation is integral to many robust geometric model fittingtechniques. Unfortunately, it is also computationally expensive, especially for higher ordergeometric models and heavily contaminated data. We propose a fundamentally new approachto accelerate hypothesis sampling by guiding it with information derived from residual sorting.We show that residual sorting innately encodes the probability of two points having arisen fromthe same model, and is obtained without recourse to domain knowledge (e.g., keypointmatching scores) typically used in previous sampling enhancement methods. More crucially,our approach encourages sampling within coherent structures and thus can very rapidlygenerate all-inlier minimal subsets that maximize the robust criterion. Sampling withincoherent structures also affords a natural ability to handle multistructure data, a condition thatis usually detrimental to other methods. The result is a sampling scheme that offers substantialspeed-ups on common computer vision tasks such as homography and fundamental matrixestimation. We show on many computer vision data, especially those with multiple structures,that ours is the only method capable of retrieving satisfactory results within realistic timebudgets.

Index Terms: Geometric model fitting, robust estimation, hypothesis generation, residualsorting, multiple structures.

325. BM3D Frames and Variational Image Deblurring.

Abstract: A family of the block matching 3-D (BM3D) algorithms for various imaging problemshas been recently proposed within the framework of nonlocal patchwise image modeling [1],[2]. In this paper, we construct analysis and synthesis frames, formalizing BM3D imagemodeling, and use these frames to develop novel iterative deblurring algorithms. We considertwo different formulations of the deblurring problem, i.e., one given by the minimization of thesingle-objective function and another based on the generalized Nash equilibrium (GNE) balanceof two objective functions. The latter results in the algorithm where deblurring and denoisingoperations are decoupled. The convergence of the developed algorithms is proved. Simulationexperiments show that the decoupled algorithm derived from the GNE formulationdemonstrates the best numerical and visual results and shows superiority with respect to thestate of the art in the field, confirming a valuable potential of BM3D-frames as an advancedimage modeling tool.

Index Terms: Deblurring, frames, image modeling, image reconstruction, sparserepresentations.

326. Re-Initialization Free Level Set Evolution Via Reaction Diffusion

Abstract: This paper presents a novel reaction-diffusion (RD) method for implicit activecontours, which is completely free of the costly re-initialization procedure in level set evolution(LSE). A diffusion term is introduced into LSE, resulting in a RD-LSE equation, to which a



piecewise constant solution can be derived. In order to have a stable numerical solution of theRD based LSE, we propose a two-step splitting method (TSSM) to iteratively solve the RD-LSEequation: first iterating the LSE equation, and then solving the diffusion equation. The secondstep regularizes the level set function obtained in the first step to ensure stability, and thus thecomplex and costly re-initialization procedure is completely eliminated from LSE. Bysuccessfully applying diffusion to LSE, the RD-LSE model is stable by means of the simple finitedifference method, which is very easy to implement. The proposed RD method can begeneralized to solve the LSE for both variational level set method and PDE-based level setmethod. The RD-LSE method shows very good performance on boundary anti-leakage, and itcan be readily extended to high dimensional level set method. The extensive and promisingexperimental results on synthetic and real images validate the effectiveness of the proposedRD-LSE approach.

Index Terms: Level set, reaction-diffusion, active contours, image segmentation, PDE,variational method

327. Monogenic Binary Coding: An Efficient Local Feature Extraction Approach To FaceRecognition

Abstract: Local feature based face recognition (FR) methods, such as Gabor features encodedby local binary pattern, could achieve state-of-the-art FR results in large-scale face databasessuch as FERET and FRGC. However, the time and space complexity of Gabor transformation aretoo high for many practical FR applications. In this paper, we propose a new and efficient localfeature extraction scheme, namely monogenic binary coding (MBC), for face representationand recognition. Monogenic signal representation decomposes an original signal into threecomplementary components: amplitude, orientation and phase. We encode the monogenicvariation in each local region and monogenic feature in each pixel, and then calculatebthestatistical features (e.g., histogram) of the extracted local features. The local statistical featuresextracted from the complementary monogenic components (i.e., amplitude, orientation andphase) are then fused for effective FR. It is shown that the proposed MBC scheme hassignificantly lower time and space complexity than the Gabor-transformation based localfeature methods. The extensive FR experiments on four large scale databases demonstratedthe effectiveness of MBC, whose performance is competitive with and even better than state-of-the-art local feature based FR methods.

Keywords: monogenic signal analysis, monogenic binary coding, face recognition, LBP, Gaborfiltering

328. Monotonic Regression: A New Way For Correlating Subjective And Objective Ratings InImage Quality Research

Abstract: To assess the performance of image quality metrics (IQMs), some regressions, such aslogistic regression and polynomial regression, are used to correlate objective ratings withsubjective scores. However, some defects in optimality are shown in these regressions. In this



correspondence, monotonic regression (MR) is found to be an effective correlation method inthe performance assessment of IQMs. Both theoretical analysis and experimental results haveproven that MR performs better than any other regression. We believe that MR could be aneffective tool for performance assessment in the IQM research.

Index Terms: Image quality assessment, image quality metric (IQM), metric performance,monotonic regression (MR).

329. Demonstration Of Real-Time Spectrum Sensing For Cognitive Radio

Abstract: Spectrum sensing detects the availability of the radio frequency spectrum in a real-time fashion, which is essential and vital to cognitive radio. The requirement for real-timeprocessing indeed poses challenges on implementing spectrum sensing algorithms. Trade-offbetween the complexity and the effectiveness of spectrum sensing algorithms should be takeninto consideration. In this paper, a fast Fourier transform (FFT) based spectrum sensingalgorithm called FAR is introduced. It is the beauty of the algorithm that the decision variable isinsensitive to noise level. Parameter selection for the algorithm is considered as well, towardminimizing computational complexity. A small form factor (SFF) software defined radio (SDR)development platform (DP) is employed to implement a spectrum sensing receiver with FARalgorithm. Performance of FAR algorithm is evaluated on the SFF SDR DP, and real-timespectrum sensing is demonstrated. FAR algorithm is friendly to hardware implementation and itis effective to detect signals at low SNR.

330. ML Estimation Of Time And Frequency Offset In OFDM Systems

Abstract: We present the joint maximum likelihood (ML) symbol-time and carrier-frequencyoffset estimator in orthogonal frequency-division multiplexing (OFDM) systems. Redundantinformation contained within the cyclic prefix enables this estimation without additional pilots.Simulations show that the frequency estimator may be used in a tracking mode and the timeestimator in an acquisition mode.

331. Efficient Encoding Of Low-Density Parity-Check Codes

Abstract: Low-density parity-check (LDPC) codes can be considered serious competitors toturbo codes in terms of performance and complexity and they are based on a similarphilosophy: constrained random code ensembles and iterative decoding algorithms. In thispaper, we consider the encoding problem for LDPC codes. More generally, we consider theencoding problem for codes specified by sparse parity-check matrices. We show how to exploitthe sparseness of the parity-check matrix to obtain efficient encoders. For the (3 6)-regularLDPC code, for example, the complexity of encoding is essentially quadratic in the block length.However, we show that the associated coefficient can be made quite small, so that encodingcodes even of length 100 000 is still quite practical. More importantly, we will show that“optimized” codes actually admit linear time encoding.



Index Terms: Binary erasure channel, decoding, encoding, parity check, random graphs, sparsematrices, turbo codes.

332. Multi-User Diversity Vs. Accurate Channel State Information In MIMO Downlink

Abstract: In a multiple transmit antenna, single antenna per receiver downlink channel withlimited channel state feedback, we consider the following question: given a constraint on thetotal system-wide feedback load, is it preferable to get low-rate/coarse channel feedback froma large number of receivers or high-rate/high-quality feedback from a smaller number ofreceivers? Acquiring feedback from many receivers allows multi-user diversity to be exploited,while high-rate feedback allows for very precise selection of beam forming directions. We showthat there is a strong preference for obtaining high-quality feedback, and that obtaining near-perfect channel information from as many receivers as possible provides a significantly largersum rate than collecting a few feedback bits from a large number of users.

333. Sum Power Iterative Water-Filling For Multi-Antenna Gaussian Broadcast Channels

Abstract: In this correspondence, we consider the problem of maximizing sum rate of amultiple-antenna Gaussian broadcast channel (BC). It was recently found that dirty-papercoding is capacity achieving for this channel. In order to achieve capacity, the optimaltransmission policy (i.e., the optimal transmit covariance structure) given the channelconditions and power constraint must be found. However, obtaining the optimal transmissionpolicy when employing dirty-paper coding is a computationally complex non convex problem.We use duality to transform this problem into a well-structured convex multiple-accesschannel (MAC) problem. We exploit the structure of this problem and derive simple and fastiterative algorithms that provide the optimum transmission policies for the MAC, which caneasily be mapped to the optimal BC policies.

Index Terms: Broadcast channel, dirty-paper coding, duality, multipleaccess channel (MAC),multiple-input multiple-output (MIMO), systems.

334. On Optimal Power Control For Delay-Constrained Communication Over Fading Channels

Abstract: In this paper, we study the problem of optimal power control for delay-constrainedcommunication over fading channels. Our objective is to find a power control law thatoptimizes the link layer performance, specifically, minimizes delay bound violation probability(or equivalently, the packet drop probability), subject to constraints on average power, arrivalrate, and delay bound. The transmission buffer size is assumed to be finite; hence, when thebuffer is full, there will be packet drop. The fading channel under our study has a continuousstate, e.g., Rayleigh fading. Since directly solving the power control problem (which optimizesthe link layer performance) is particularly challenging, we decompose it into three subproblems, and solve the three sub-problems iteratively; we call the resulting scheme JointQueue Length Aware (JQLA) power control, which produces a local optimal solution to thethree sub problems. We prove that the solution that simultaneously solves the three sub-



problems is also an optimal solution to the optimal power control problem. Simulation resultsshow that the JQLA scheme achieves superior performance over the time domain water fillingand the truncated channel inversion power control. E.g., JQLA achieves 10 dB gain at packetdrop probability of 10¡3, over the time domain water filling power control.

Index Terms: Delay-constrained communication, power control, queuing analysis, delay boundviolation probability, packet drop probability.

336. An Improved Algorithm For Blind Reverberation Time Estimation

Abstract: An improved algorithm for the estimation of the reverberation time (RT) fromreverberant speech signals is presented. This blind estimation of the RT is based on a simplestatistical model for the sound decay such that the RT can be estimated by means of amaximum-likelihood (ML) estimator. The proposed algorithm has a significantly lowercomputational complexity than previous ML-based algorithms for RT estimation. This isachieved by a down sampling operation and a simple pre-selection of possible sound decays.The new algorithm is more suitable to track time-varying RTs than related approaches. Inaddition, it can also estimate the RT in the presence of (moderate) background noise. Theproposed algorithm can be employed to measure the RT of rooms from sound recordingswithout using a dedicated measurement setup. Another possible application is its use withinspeech de reverberation systems for hands-free devices or digital hearing aids.

Index Terms: reverberation time, blind estimation, low complexity, speech dereverberation

337. Fast And Accurate Sequential Floating Forward Feature Selection With the BayesClassifier Applied To Speech Emotion Recognition

Abstract: This paper addresses subset feature selection performed by the sequential floatingforward selection (SFFS). The criterion employed in SFFS is the correct classification rate of theBayes classifier assuming that the features obey the multivariate Gaussian distribution. Atheoretical analysis that models the number of correctly classified utterances as a hypergeometric random variable enables the derivation of an accurate estimate of the variance ofthe correct classification rate during cross-validation. By employing such variance estimate, wepropose a fast SFFS variant. Experimental findings on Danish emotional speech (DES) andSpeech Under Simulated and Actual Stress (SUSAS) databases demonstrate that SFFScomputational time is reduced by 50% and the correct classification rate for classifying speechinto emotional states for the selected subset of features varies less than the correctclassification rate found by the standard SFFS. Although the proposed SFFS variant is tested inthe framework of speech emotion recognition, the theoretical results are valid for any classifierin the context of any wrapper algorithm.

Key words: Bayes classifier, cross-validation, variance of the correct classification rate of theBayes classifier, feature selection, wrappers



338. Hybrid De Algorithm With Adaptive Crossover Operator For Solving Real-WorldNumerical Optimization Problems

Abstract: In this paper, the results for the CEC 2011 Competition on testing evolutionaryalgorithms on real world optimization problems using a hybrid differential evolution algorithmare presented. The proposal uses a local search routine to improve convergence and anadaptive crossover operator. According to the obtained results, this algorithm shows to be ableto find competitive solutions with reported results.

Index Terms: Differential Evolution algorithm, parameter selection, CEC competition.

339. Real-Time Compressive Tracking

Abstract: It is a challenging task to develop effective and efficient appearance models forrobust object tracking due to factors such as pose variation, illumination change, occlusion, andmotion blur. Existing online tracking algorithms often update models with samples fromobservations in recent frames. While much success has been demonstrated, numerous issuesremain to be addressed. First, while these adaptive appearance models are data-dependent,there does not exist sufficient amount of data for online algorithms to learn at the outset.Second, online tracking algorithms often encounter the drift problems. As a result of self-taughtlearning, these mis-aligned samples are likely to be added and degrade the appearance models.In this paper, we propose a simple yet effective and efficient tracking algorithm with anappearance model based on features extracted from the multi-scale image feature space withdata-independent basis. Our appearance model employs nonadaptive random projections thatpreserve the structure of the image feature space of objects. A very sparse measurementmatrix is adopted to efficiently extract the features for the appearance model. We compresssamples of foreground targets and the background using the same sparse measurement matrix.The tracking task is formulated as a binary classification via a naive Bayes classifier with onlineupdate in the compressed domain. The proposed compressive tracking algorithm runs in real-time and performs favorably against state-of-the-art algorithms on challenging sequences interms of efficiency, accuracy and robustness.

340. An Efficient Algorithm For Level Set Method Preserving Distance Function

Abstract: The level set method is a popular technique for tracking moving interfaces in severaldisciplines including computer vision and fluid dynamics. However, despite its high flexibility,the original level set method is limited by two important numerical issues. Firstly, the level setmethod does not implicitly preserve the level set function as a distance function, which isnecessary to estimate accurately geometric features s.a. the curvature or the contour normal.Secondly, the level set algorithm is slow because the time step is limited by the standard CFLcondition, which is also essential to the numerical stability of the iterative scheme. Recentadvances with graph cut methods and continuous convex relaxation provide powerfulalternatives to the level set method for image processing problems because they are fast,accurate and guaranteed to find the global minimizer independently to the initialization. These



recent techniques use binary functions to represent the contour rather than distance functions,which are usually considered for the level set method. However, the binary function cannotprovide the distance information, which can be essential for some applications s.a. the surfacereconstruction problem from scattered points and the cortex segmentation problem in medicalimaging. In this paper, we propose a fast algorithm to preserve distance functions in level setmethods. Our algorithm is inspired by recent efficient `1 optimization techniques, which willprovide an efficient and easy to implement algorithm. It is interesting to note that ouralgorithm is not limited by the CFL condition and it naturally preserves the level set function asa distance function during the evolution, which avoids the classical re-distancing problem inlevel set methods. We apply the proposed algorithm to carry out image segmentation, whereour methods proves to be 5 to 6 times faster than standard distance preserving level settechniques. We also present two applications where preserving a distance function is essential.Nonetheless, our method stays generic and can be applied to any level set methods thatrequire the distance information.

Index Terms: Level set, image segmentation, surface reconstruction, signed distance function,numerical scheme, splitting.

341. Efficient Misalignment-Robust Representation For Real-Time Face Recognition

Abstract: Sparse representation techniques for robust face recognition have been widelystudied in the past several years. Recently face recognition with simultaneous misalignment,occlusion and other variations has achieved interesting results via robust alignment by sparserepresentation (RASR). In RASR, the best alignment of a testing sample is sought subject bysubject in the database. However, such an exhaustive search strategy can make the timecomplexity of RASR prohibitive in large-scale face databases. In this paper, we propose a novelscheme, namely misalignment robust representation (MRR), by representing the misalignedtesting sample in the transformed face space spanned by all subjects. The MRR seeks the bestalignment via a two-step optimization with a coarse-to-fine search strategy, which needs onlytwo deformation-recovery operations. Extensive experiments on representative face databasesshow that MRR has almost the same accuracy as RASR in various face recognition andverification tasks but it runs tens to hundreds of times faster than RASR. The running time ofMRR is less than 1 second in the large-scale Multi-PIE face database, demonstrating its greatpotential for real-time face recognition.

342. Robust Point Matching Revisited: A Concave Optimization Approach

Abstract: The well-known robust point matching (RPM) method uses deterministic annealingfor optimization, and it has two problems. First, it cannot guarantee the global optimality of thesolution and tends to align the centers of two point sets. Second, deformation needs to beregularized to avoid the generation of undesirable results. To address these problems, in thispaper we show that the energy function of RPM can be reduced to a concave function withvery few non-rigid terms after eliminating the transformation variables and applying lineartransformation; we then propose to use concave optimization technique to minimize the



resulting energy function. The proposed method scales well with problem size, achieves theglobally optimal solution, and does not need regularization for simple transformations such assimilarity transform. Experiments on synthetic and real data validate the advantages of ourmethod in comparison with state-of-the-art methods.

343. Canny Edge Detection Enhancement By Scale Multiplication

344. Robust Object Tracking Using Joint Color-Texture Histogram

Abstract: A novel object tracking algorithm is presented in this paper by using the joint colortexture histogram to represent a target and then applying it to the mean shift framework. Apartfrom the conventional color histogram features, the texture features of the object are alsoextracted by using the local binary pattern (LBP) technique to represent the object. The majoruniform LBP patterns are exploited to form a mask for joint color-texture feature selection.Compared with the traditional color histogram based algorithms that use the whole targetregion for tracking, the proposed algorithm extracts effectively the edge and corner features inthe target region, which characterize better and represent more robustly the target. Theexperimental results validate that the proposed method improves greatly the tracking accuracyand efficiency with fewer mean shift iterations than standard mean shift tracking. It canrobustly track the target under complex scenes, such as similar target and backgroundappearance, on which the traditional color based schemes may fail to track.

Keywords: Object tracking; mean shift; local binary pattern; color histogram.

345. Distance Regularized Level Set Evolution And Its Application To Image Segmentation

Abstract: Level set methods have been widely used in image processing and computer vision. Inconventional level set formulations, the level set function typically develops irregularitiesduring its evolution, which may cause numerical errors and eventually destroy the stability ofthe evolution. Therefore, a numerical remedy, called re initialization, is typically applied toperiodically replace the degraded level set function with a signed distance function. However,



the practice of re initialization not only raises serious problems as when and how it should beperformed, but also affects numerical accuracy in an undesirable way. This paper proposes anew variational level set formulation in which the regularity of the level set function isintrinsically maintained during the level set evolution. The level set evolution is derived as thegradient flow that minimizes an energy functional with a distance regularization term and anexternal energy that drives the motion of the zero level set toward desired locations. Thedistance regularization term is defined with a potential function such that the derived level setevolution has a unique forward-and-backward (FAB) diffusion effect, which is able to maintain adesired shape of the level set function, particularly a signed distance profile near the zero levelset. This yields a new type of level set evolution called distance regularized level set evolution(DRLSE). The distance regularization effect eliminates the need for reinitialization and therebyavoids its induced numerical errors. In contrast to complicated implementations ofconventional level set formulations, a simpler and more efficient finite difference scheme canbe used to implement the DRLSE formulation. DRLSE also allows the use of more general andefficient initialization of the level set function. In its numerical implementation, relatively largetime steps can be used in the finite difference scheme to reduce the number of iterations, whileensuring sufficient numerical accuracy. To demonstrate the effectiveness of the DRLSEformulation, we apply it to an edge-based active contour model for image segmentation, andprovide a simple narrowband implementation to greatly reduce computational cost.

Index Terms: Forward and backward diffusion, image segmentation, level set method,narrowband, reinitialization.

346. Minimization Of Region-Scalable Fitting Energy For Image Segmentation

Abstract: Intensity inhomogeneities often occur in real-world images and may causeconsiderable difficulties in image segmentation. In order to overcome the difficulties caused byintensity inhomogeneities, we propose a region-based active contour model that draws uponintensity information in local regions at a controllable scale. A data fitting energy is defined interms of a contour and two fitting functions that locally approximate the image intensities onthe two sides of the contour. This energy is then incorporated into a variational level setformulation with a level set regularization term, from which a curve evolution equation isderived for energy minimization. Due to a kernel function in the data fitting term, intensityinformation in local regions is extracted to guide the motion of the contour, which therebyenables our model to cope with intensity inhomogeneity. In addition, the regularity of the levelset function is intrinsically preserved by the level set regularization term to ensure accuratecomputation and avoids expensive reinitialization of the evolving level set function.Experimental results for synthetic and real images show desirable performances of ourmethod.

Index Terms: Image segmentation, intensity inhomogeneity, level set method, region-scalablefitting energy, variational method.



347. Motion Tracking

Abstract: The motion tracking task was decomposed into two independent subproblems. The_rst is to detect foreground objects on a frame-wise basis, by labelling each pixel in an imageframe as either foreground or background. The second is to couple object observations atdi_erent points in a sequence to yield the object's motion trajectory.

348. A Level Set Method For Image Segmentation In The Presence Of Intensity Inhomogeneities With Application To MRI

Abstract: Intensity inhomogeneity often occurs in real-world images, which presents aconsiderable challenge in image segmentation. The most widely used image segmentationalgorithms are region-based and typically rely on the homogeneity of the image intensities inthe regions of interest, which often fail to provide accurate segmentation results due to theintensity inhomogeneity. This paper proposes a novel region-based method for imagesegmentation, which is able to deal with intensity inhomogeneities in the segmentation. First,based on the model of images with intensity inhomogeneities, we derive a local intensityclustering property of the image intensities, and define a local clustering criterion function forthe image intensities in a neighborhood of each point. This local clustering criterion function isthen integrated with respect to the neighborhood center to give a global criterion of imagesegmentation. In a level set formulation, this criterion defines an energy in terms of the levelset functions that represent a partition of the image domain and a bias field that accounts forthe intensity inhomogeneity of the image. Therefore, by minimizing this energy, our method isable to simultaneously segment the image and estimate the bias field, and the estimated biasfield can be used for intensity inhomogeneity correction (or bias correction). Our method hasbeen validated on synthetic images and real images of various modalities, with desirableperformance in the presence of intensity inhomogeneities. Experiments show that our methodis more robust to initialization, faster and more accurate than the well-known piecewisesmooth model. As an application, our method has been used for segmentation and biascorrection of magnetic resonance (MR) images with promising results.

Index Terms: Bias correction, image segmentation, intensity inhomogeneity, level set, MRI.

164782623-Matlab-2013-pdf

Documents

Transcript of 164782623-Matlab-2013-pdf