Recently, sparsity based classification has been applied to video anomaly detection. A linear model is assumed over video features (e.g. trajectories) such that the feature representation of a new event is written as a sparse linear combination of existing feature representations in the dictionary. Sparsity based video anomaly detection shows promise but open challenges remain in that existing methods assume object specific and class specific event dictionaries making them applicable mostly in highly structured scenarios. Second, using conventional sparsity models on matrices/vectors, the computational burden is often high. In this work, we advocate a more general and practical sparsity model using a low-rank structure on the matrix of sparse coefficients. We find that enforcing a low-rank structure can ease the rigidity of traditional row-sparse constraints on sparse coefficient vectors/matrices. Because low-rank matrices are of course not always sparse, an additional l1regularization term is added. Further, if rank is substituted by its convex nuclear norm alternative, then significant computational benefits can be obtained over existing methods in sparsity based video anomaly detection. Experimental evaluation on benchmark video datasets reveal, our method is competitive with state-of-the art while providing robustness benefits under occlusion.

Video anomaly detection can be used in the transportation domain to identify unusual patterns such as traffic violations, accidents, unsafe driver behavior, street crime, and other suspicious activities. A common class of approaches relies on object tracking and trajectory analysis. Very recently, sparse reconstruction techniques have been employed in video anomaly detection. The fundamental underlying assumption of these methods is that any new feature representation of a normal/anomalous event can be approximately modeled as a (sparse) linear combination prelabeled feature representations (of previously observed events) in a training dictionary. Sparsity can be a powerful prior on model coefficients but challenges remain in the detection of anomalies involving multiple objects and the ability of the linear sparsity model to effectively allow for class separation. The proposed research addresses both these issues. First, we develop a new joint sparsity model for anomaly detection that enables the detection of joint anomalies involving multiple objects. This extension is highly nontrivial since it leads to a new simultaneous sparsity problem that we solve using a greedy pursuit technique. Second, we introduce nonlinearity into, that is, kernelize. The linear sparsity model to enable superior class separability and hence anomaly detection. We extensively test on several real world video datasets involving both single and multiple object anomalies. Results show marked improvements in detection of anomalies in both supervised and unsupervised scenarios when using the proposed sparsity models.

A new framework is proposed to embed watermark through alternating frequency components of the black plane of a printed CMYK image according to the watermark message, and then decode the watermark with an infrared (IR) scanner. This framework leverages the facts that a given color can be produced by multiple combinations of CMYK quadruplets and the standard four-color printing materials have very different infrared transmission characteristics. Carbon-based black toner is a good absorber of infrared light, while the other color toners are almost transparent in the infrared region. In the proposed framework, a watermark message is first embedded in the black plane. After that, a color adjustment step is followed to modify the CMY planes in order to minimize visibility of the watermark. Thus, the proposed method has the advantage of low visual noise caused by watermark embedding. In addition, it is available for “blind decoding”.

The JPEG standard is one of the most prevalent image compression schemes in use today. While JPEG was designed for use with natural images, it is also widely used for the encoding of raster documents. Unfortunately, JPEG's characteristic blocking and ringing artifacts can severely degrade the quality of text and graphics in complex documents. We propose a JPEG decompression algorithm which is designed to produce substantially higher quality images from the same standard JPEG encodings. The method works by incorporating a document image model into the decoding process which accounts for the wide variety of content in modern complex color documents. The method works by first segmenting the JPEG encoded document into regions corresponding to background, text, and picture content. The regions corresponding to text and background are then decoded using maximum a posteriori (MAP) estimation. Most importantly, the MAP reconstruction of the text regions uses a model which accounts for the spatial characteristics of text and graphics. Our experimental comparisons to the baseline JPEG decoding as well as to three other decoding schemes, demonstrate that our method substantially improves the quality of decoded images, both visually and as measured by PSNR.

We routinely encounter digital color images that were previously compressed using the Joint Photographic Experts Group (JPEG) standard. En route to the image's current representation, the previous JPEG compression's various settings-termed its JPEG compression history (CH)-are often discarded after the JPEG decompression step. Given a JPEG-decompressed color image, this paper aims to estimate its lost JPEG CH. We observe that the previous JPEG compression's quantization step introduces a lattice structure in the discrete cosine transform (DCT) domain. This paper proposes two approaches that exploit this structure to solve the JPEG Compression History Estimation (CHEst) problem. First, we design a statistical dictionary-based CHEst algorithm that tests the various CHs in a dictionary and selects the maximum a posteriori estimate. Second, for cases where the DCT coefficients closely conform to a 3-D parallelepiped lattice, we design a blind lattice-based CHEst algorithm. The blind algorithm exploits the fact that the JPEG CH is encoded in the nearly orthogonal bases for the 3-D lattice and employs novel lattice algorithms and recent results on nearly orthogonal lattice bases to estimate the CH. Both algorithms provide robust JPEG CHEst performance in practice. Simulations demonstrate that JPEG CHEst can be useful in JPEG recompression; the estimated CH allows us to recompress a JPEG-decompressed image with minimal distortion (large signal-to-noise-ratio) and simultaneously achieve a small file-size.

We routinely encounter digital color images that were previously JPEG-compressed. We aim to retrieve the various settings - termed JPEG compression history (CH) - employed during previous JPEG operations. This information is often discarded en-route to the image's current representation. The discrete cosine transform coefficient histograms of previously JPEG-compressed images exhibit near-periodic behavior due to quantization. We propose a statistical approach to exploit this structure and thereby estimate the image's CH. Using simulations, we first demonstrate the accuracy of our estimation. Further, we show that JPEG recompression performed by exploiting the estimated CH strikes an excellent file-size versus distortion tradeoff.

Sometimes image processing units inherit images in raster bitmap format only, so that processing is to be carried without knowledge of past operations that may compromise image quality (e.g., compression). To carry further processing, it is useful to not only know whether the image has been previously JPEG compressed, but to learn what quantization table was used. This is the case, for example, if one wants to remove JPEG artifacts or for JPEG re-compression. In this paper, a fast and efficient method is provided to determine whether an image has been previously JPEG compressed. After detecting a compression signature, we estimate compression parameters. Specifically, we developed a method for the maximum likelihood estimation of JPEG quantization steps. The quantizer estimation method is very robust so that only sporadically an estimated quantizer step size is off, and when so, it is by one value.

We propose a segmentation algorithm for color document images called background identification based segmentation (BISeg). This algorithm efficiently locates and classifies objects in an image by identifying both the main background (paper background) and local backgrounds. In addition, BISeg generates a tree representation of document images, and the resulting segmentation can be used for many document processing applications including enhancement, rendering, editing and compression.

High-level (semantic) image classification can be achieved by analysis of low-level image attributes geared for the particular classes. In this paper, we have proposed a novel application of the known image processing and classification techniques to achieve such a high-level classification of color images. Our image classification algorithm uses three low-level image features: texture, color, and edge characteristics to classify a color image into two classes: business graphics or natural picture. We have achieved an accuracy of 96.6% on our database of 209 images using a combination of tree and neural network classifiers.

To process previously JPEG coded images the knowledge of the quantization table used in compression is sometimes required. This happens for example in JPEG artifact removal and in JPEG re-compression. However, the quantization table might not be known due to various reasons. A method is presented for the maximum likelihood estimation (MLE) of the JPEG quantization tables. An efficient method is also provided to identify if an image has been previously JPEG compressed.

Compound document images contain graphic or textual content along with pictures. They are a very common form of documents, found in magazines, brochures, Web-sites, etc. We focus our attention on the mixed raster content (MRC) multi-layer approach for compound image compression. We study block thresholding as a means to segment an image for MRC. An attempt is made to optimize the block-threshold in a rate-distortion sense. Rate-distortion curves are presented to demonstrate the performance of the proposed algorithm.

Compound document images contain graphic or textual content along with pictures. They are a very common form of documents, found in magazines, brochures, Web sites, etc. We focus our attention on the mixed raster content (MRC) multilayer approach for compound image compression. We study block thresholding as a means to segment an image for MRC. An attempt is made to optimize the block threshold in a rate-distortion sense. Also, a fast algorithm is presented to approximate the optimized method. Extensive results are presented including rate-distortion curves, segmentation masks and reconstructed images, showing the performance of the proposed algorithm.

Block-based discrete cosine transform (BDCT) image coding techniques are widely used in image and video compression applications such as JPEG and MPEG. At a moderate bit rate, the BDCT is usually a quite satisfactory solution to most of practical coding applications. However, for high compression rates it produces noticeable blocking and ringing artifacts in the decompressed image. We propose a novel post-processing algorithm to remove the blocking and ringing effects at low bit rates. The main steps in this algorithm include block classification, boundary low-pass filtering and mid-point interpolation, edge detection and filtering, and DCT coefficient constraint. The improvement is demonstrated both subjectively and objectively.

This paper presents a method for retrieving a gray picture from a halftone image. The algorithm performs a two dimensional (2-D) filtering which is composed of a 1-D lowpass filtering in one direction and a 1-D pattern matching in a perpendicular direction. It provides substantially better detail than the traditional lowpass filtering approach, yet it is robust to noise and distortions. The technique can be applied to both orthographical halftones and scanned halftones.

The major artifacts of the JPEG compressed images are blocking and ringing, which are mainly due to the coarse quantization of low frequency and high frequency DCT components, respectively. At a very low bit rate, these artifacts might be rather disturbing to human eyes. While research for reduction of blocking effects has been relatively extensively conducted, little has been done for improving ringing. The latter is particularly dominant in document images, where sharp edges present commonly in text and graphics are more likely to be encountered. In this paper we propose an algorithm, which was designed mainly for ringing effects removal. The same principle can be also applied to reduce blocking effects.

Adaptive equalization can be used to improve digital data transmission on wireless links with time-varying multipath distortion. It was proved by Zhou, Proakis and Ling (see IEEE Transactions on Communications, vol.38, p.8-24, no.1, 1990) that feedforward schemes are universally capable of approximating any measurable function to any desired degree of accuracy. We propose a new realization of such feedforward scheme to the channel equalization problem, parallel feedforward equalization (PFE). An important feature of the new approach is the decomposition of any equalization into linear and nonlinear components. The new approach chooses F/sub j/(/spl middot/) (j=1, ...) from a family of nonlinear functions to approximate the nonlinear component decomposed from the desired mapping f(/spl middot/). The other new idea proposed in this paper is a measure called nonlinearity distribution which characterizes the nonlinearity in multipath fading channels. The architecture of the new equalization consists of parallel feedforward nonlinear filters, each of them has a specifically tailored nonlinear function F/sub j/(/spl middot/).

Error diffusion is an important technique for digital halftoning. One of the most observable artifacts of error diffusion is the inherent edge enhancement. We analyze the edge behavior of error diffusion. We show that error diffusion does not always enhance edges. For some error diffusion weights, it actually blurs the images. We demonstrate that the source for edge enhancement or blurring is the nonuniform distribution of the correction terms. We propose an enhancement index to predict the overall edge behavior for a given set of weights.

Error diffusion is an important technique for digital halftoning. One of the most observable artifacts of error diffusion is the periodically repeating patterns appearing in the areas where the input is constant or slowly varying. We analyze the limit cycle behavior of error diffusion, which explains the periodic oscillation under a constant input. We discuss why certain periodic patterns are more likely to appear than the others in the image. We demonstrate how to increase or reduce the occurrence of a pattern by manipulating the error diffusion weights.>

The authors discuss the problem of textile fabric inspection using the visual textural properties of the fabric. The problem is to detect and locate the various kinds of defects that might be present in a given fabric sample based on an image of the fabric. Stochastic models are used to model the visual fabric texture. The authors use the Gaussian Markov random field to model the texture image of nondefective fabric. The inspection problem is cast as a statistical hypothesis testing problem on statistics derived from the model. The image of the fabric patch to be inspected is partitioned into nonoverlapping windows of size N*N where each window is classified as defective or nondefective based on a likelihood ratio test of size alpha . The test is recast in terms of the sufficient statistics associated with the model parameters. The sufficient statistics are easily computable for any sample. The authors generalize the test when the model parameters of the fabric are assumed to be unknown.>

Consideration is given to the problem of classifying a test textured image that is obtained from one of C possible parent texture classes, after possibly applying unknown rotation and scale changes to the parent texture. The training texture images (parent classes) are modeled by Gaussian Markov random fields (GMRFs). To classify a rotated and scaled test texture, the rotation and scale changes are incorporated in the texture model through an appropriate transformation of the power spectral density of the GMRF. For the rotated and scaled image, a bona fide likelihood function that shows the explicit dependence of the likelihood function on the GMRF parameters, as well as on the rotation and scale parameters, is derived. Although, in general, the scaled and/or rotated texture does not correspond to a finite-order GMRF, it is possible nonetheless to write down a likelihood function for the image data. The likelihood function of the discrete Fourier transform of the image data corresponds to that of a white nonstationary Gaussian random field, with the variance at each pixel (i,j) being a known function of the rotation, the scale, the GMRF model parameters, and (i,j). The variance is an explicit function of the appropriately sampled power spectral density of the GMRF. The estimation of the rotation and scale parameters is performed in the frequency domain by maximizing the likelihood function associated with the discrete Fourier transform of the image data. Cramer-Rao error bounds on the scale and rotation estimates are easily computed. A modified Bayes decision rule is used to classify a given test image into one of C possible texture classes. The classification power of the method is demonstrated through experimental results on natural textures from the Brodatz album.>

An edge-based hierarchical algorithm is proposed for segmenting textured images. The texture regions are modeled by Gaussian Markov random fields. No prior knowledge about the texture parameter values or the number of texture regions is assumed. The algorithm consists of two stages: boundary (edge) detection and edge estimation. In the first stage, the image is divided into disjoint square windows. The windows on the boundary of the regions are detected through a hypothesis test. The generalized likelihood ratio (GLR) shows some asymptotic optimality for the test. To reduce the computational cost associated with the GLR, average periodogram and low rank approximation are applied. The exact locations of the edges are hierarchically estimated in the second stage by a maximum likelihood estimator. Simulation and experimental examples are discussed.>

The problem of classifying a textured image which might be subject to unknown rotation and magnification scale changes into one of C possible texture classes is discussed. The texture classes are modeled by Gaussian Markov random fields. A Bayes decision rule based on the generalized likelihood function is used to classify a given test sample. A maximum-likelihood estimate for the scale and rotation parameters for each of the C texture classes is computed under the assumption that the observed texture came from a particular unrotated and unscaled texture model. The test texture is allocated to the class with the highest generalized likelihood function. The classification power of the method is demonstrated through extensive experimental results on natural texture from the Brodatz album as well as for the problem of fabric inspection.>