Abstract: Discriminative methods for visual object category recognition are typically non-probabilistic, predicting class labels but not directly providing an estimate of uncertainty. Gaussian Processes (GPs) are powerful regression techniques with explicit uncertainty models; we show here how Gaussian Processes with covariance functions defined based on a Pyramid Match Kernel (PMK) can be used for probabilistic object category recognition. The uncertainty model provided by GPs offers confidence estimates at test points, and naturally allows for an active learning paradigm in which points are optimally selected for interactive labeling. We derive a novel active category learning method based on our probabilistic regression model, and show that a significant boost in classification performance is possible, especially when the amount of training data for a category is ultimately very small.

Abstract: In 1992, Bamberger and Smith proposed the directional filter bank (DFB) for an efficient directional decomposition of 2-D signals. Due to the nonseparable nature of the system, extending the DFB to higher dimensions while still retaining its attractive features is a challenging and previously unsolved problem. We propose a new family of filter banks, named NDFB, that can achieve the directional decomposition of arbitrary N-dimensional (Nges2) signals with a simple and efficient tree-structured construction. In 3-D, the ideal passbands of the proposed NDFB are rectangular-based pyramids radiating out from the origin at different orientations and tiling the entire frequency space. The proposed NDFB achieves perfect reconstruction via an iterated filter bank with a redundancy factor of N in N-D. The angular resolution of the proposed NDFB can be iteratively refined by invoking more levels of decomposition through a simple expansion rule. By combining the NDFB with a new multiscale pyramid, we propose the surfacelet transform, which can be used to efficiently capture and represent surface-like singularities in multidimensional data

Abstract: In numerous domains it is useful to represent a single example by the set of the local features or parts that comprise it. However, this representation poses a challenge to many conventional machin...

Abstract: In the paper, we propose a robust and fast image denoising method. The approach integrates both 'Non-Local means algorithm and Laplacian Pyramid. Given an image to be denoised, we first decompose it into Laplacian pyramid. Exploiting the redundancy property of Laplacian pyramid, we then perform non-local means on every level image of Laplacian pyramid. Essentially, we use the similarity of image features in Laplacian pyramid to act as weight to denoise image. Since the features extracted in Laplacian pyramid are localized in spatial position and scale, they are much more able to describe image, and computing the similarity between them is more reasonable and more robust. Also, based on the efficient Summed Square Image (SSI) scheme and Fast Fourier Transform (FFT), we present an accelerating algorithm to break the bottleneck of non-local means algorithm -similarity computation of compare windows. After speedup, our algorithm is fifty times faster than original non-local means algorithm. Experiments demonstrated the effectiveness of our algorithm.

Abstract: Matching local features across images is often useful when comparing or recognizing objects or scenes, and efficient techniques for obtaining image-to-image correspondences have been developed [4, 11, 16]. However, given a query image, searching a very large image database with such measures remains impractical. We introduce a sub-linear time randomized hashing algorithm for indexing sets of feature vectors under their partial correspondences. We develop an efficient embedding function for the normalized partial matching similarity between sets, and show how to exploit random hyperplane properties to construct hash functions that satisfy locality-sensitive constraints. The result is a bounded approximate similarity search algorithm that finds (1 + epsiv)-approximate nearest neighbor images in O(N1/1+epsiv) time for a database containing N images represented by (varying numbers of) local features. We demonstrate our approach applied to image retrieval for images represented by sets of local appearance features, and show that searching over correspondences is now scalable to large image databases.

Abstract: Recognition of 3D objects from different viewpoints is a difficult problem. In this paper, we propose a new method to recognize 3D range images by matching local surface descriptors. The input 3D surfaces are first converted into a set of local shape descriptors computed on surface patches defined by detected salient features. We compute the similarities between input 3D images by matching their descriptors with a pyramid kernel function. The similarity matrix of the images is used to train for classification using SVM, and new images can be recognized by comparing with the training set. The approach is evaluated on both synthetic and real 3D data with complex shapes.

Abstract: A robust face detection technique along with mouth localization, processing every frame in real time (video rate), is pre- sented. Moreover, it is exploited for motion analysis onsite to verify "liveness" as well as to achieve lip reading of digits. A method- ological novelty is the suggested quantized angle features ("quan- gles") being designed for illumination invariance without the need for preprocessing (e.g., histogram equalization). This is achieved by using both the gradient direction and the double angle direc- tion (the structure tensor angle), and by ignoring the magnitude of the gradient. Boosting techniques are applied in a quantized fea- ture space. A major benefit is reduced processing time (i.e., that the training of effective cascaded classifiers is feasible in very short time, less than 1 h for data sets of order ). Scale invariance is implemented through the use of an image scale pyramid. We propose "liveness" verification barriers as applications for which a significant amount of computation is avoided when estimating mo- tion. Novel strategies to avert advanced spoofing attempts (e.g., re- played videos which include person utterances) are demonstrated. We present favorable results on face detection for the YALE face test set and competitive results for the CMU-MIT frontal face test set as well as on "liveness" verification barriers.

Abstract: In the paper, we propose a robust and fast image denoising method. The approach integrates both 'Non-Local means algorithm and Laplacian Pyramid. Given an image to be denoised, we first decompose it into Laplacian pyramid. Exploiting the redundancy property of Laplacian pyramid, we then perform non-local means on every level image of Laplacian pyramid. Essentially, we use the similarity of image features in Laplacian pyramid to act as weight to denoise image. Since the features extracted in Laplacian pyramid are localized in spatial position and scale, they are much more able to describe image, and computing the similarity between them is more reasonable and more robust. Also, based on the efficient Summed Square Image (SSI) scheme and Fast Fourier Transform (FFT), we present an accelerating algorithm to break the bottleneck of non-local means algorithm -similarity computation of compare windows. After speedup, our algorithm is fifty times faster than original non-local means algorithm. Experiments demonstrated the effectiveness of our algorithm.

Abstract: In this paper, a multiresolution volumetric texture segmentation (M-VTS) algorithm is presented. The method extracts textural measurements from the Fourier domain of the data via subband filtering using an orientation pyramid (Wilson and Spann, 1988). A novel Bhattacharyya space, based on the Bhattacharyya distance, is proposed for selecting the most discriminant measurements and producing a compact feature space. An oct tree is built of the multivariate features space and a chosen level at a lower spatial resolution is first classified. The classified voxel labels are then projected to lower levels of the tree where a boundary refinement procedure is performed with a three-dimensional (3-D) equivalent of butterfly filters. The algorithm was tested with 3-D artificial data and three magnetic resonance imaging sets of human knees with encouraging results. The regions segmented from the knees correspond to anatomical structures that can be used as a starting point for other measurements such as cartilage extraction

Abstract: We present a novel automatic multiscale algorithm applied to segmentation of anatomical structures in brain MRI. The algorithm which is derived from algebraic multigrid, uses a graph representation of the image and performs a coarsening process that produces a full hierarchy of segments. Our main contribution is the incorporation of prior knowledge information into the multiscale framework through a Bayesian formulation. The probabilistic information is based on an atlas prior and on a likelihood function estimated from a manually labeled training set. The significance of our new approach is that the constructed pyramid, reflects the prior knowledge formulated. This leads to an accurate and efficient methodology for detection of various anatomical structures simultaneously. Quantitative validation results on gold standard MRI show the benefit of our approach.

Abstract: A model of a nonmodulated pyramid wavefront sensor (P-WFS) based on Fourier optics has been presented. Linearizations of the model represented as Jacobian matrices are used to improve the P-WFS phase estimates. It has been shown in simulations that a linear approximation of the P-WFS is sufficient in closed-loop adaptive optics. Also a method to compute model-based synthetic P-WFS command matrices is shown, and its performance is compared to the conventional calibration. It was observed that in poor visibility the new calibration is better than the conventional.

Abstract: This work presents a generic two-dimensional (2-D) Gaussian smoothing filter for noise image processing. The filter includes the power-of-two approximation arithmetic algorithm for the Gaussian coefficients and effective hardware design. We hope the proposed generic Gaussian smoothing filter is able to provide various levels of noise smoothing and reduction, which are highly desired in early stages of the image processing flow. By using the power-of-two terms, the generic 2-D Gaussian filter can be implemented by using simple hardware shifters and adders. An embedded multi-access SRAM concept is also adopted to raise the hardware throughput. The synthesized hardware of such combination exhibits that the proposed filter can attain a real-time high frame-rate and high-resolution video processing.

Abstract: This paper describes a rarity-based visual attention model working on both still images and video sequences. Applications of this kind of models are numerous and we focus on a perceptual image representation which enhances the perceptually important areas and uses lower resolution for perceptually less important regions. Our aim is to provide an approximation of human perception by visualizing its gradual discovery of the visual environment. Comparisons with classical methods for visual attention show that the proposed algorithm is well adapted to anisotropic filtering purposes. Moreover, it has a high ability to preserve perceptually important areas as defects or abnormalities from an important loss of information. High accuracy on low-contrast defects and scalable real-time video compression may be some practical applications of the proposed image representation.

Abstract: The present invention is to provide a denoise method on Gaussian/Laplacian image pyramid, which integrates Pyramid analysis/synthesis algorithm, MMSE (minimum mean square error) filter and NL (non local) filter on the image pyramid to reconstruct and output a denoised image of an original input image through a plurality of iterative procedures, and utilizes an auto-adaptive noise estimation algorithm to find parameter of noise level used by the NL filter, so as to be easily embedded in mobile or handheld devices for obtaining better noise removing and anti-shaking results and remove noise much faster than the conventional denoise method, but only with less quality loss.

Abstract: In this paper we discuss the significance of representation of images for face verification. We consider three different representations, namely, edge gradient, edge orientation and potential field derived from the edge gradient. These representations are examined in the context of face verification using a specific type of correlation filter, called the minimum average correlation energy (MACE) filter. The different representations are derived using one-dimensional (1-D) processing of image. The 1-D processing provides multiple partial evidences for a given face image, one evidence for each direction of the 1-D processing. Separate MACE filters are used for deriving each partial evidence. We propose a method to combine the partial evidences obtained for each representation using an auto-associative neural network (AANN) model, to arrive at a decision for face verification. Results show that the performance of the system using potential field representation is better than that using the edge gradient representation or the edge orientation representation. Also, the potential field representation derived from the edge gradient is observed to be less sensitive to variation in illumination compared to the gray level representation of images.

Abstract: This paper presents a cascaded system for recognition of the circular road-signs. The system consists of two compound detectors-classifiers. Each operates on the Gaussian scale-space and does template matching in the log-polar domain. The first module is responsible for detection of the potential sign areas at the coarsest level of the pyramid. The second one, in turn, refines the already found places at the finest level. Thanks to this composition, as well as to the efficient matching in the log-polar domain, the system is very robust in terms of recognition of the signs with different scales and rotations, as well as under partial occlusions, poor illumination conditions, and noise.

Abstract: With the concept of the packing problem,this paper presents a color image representation method based on the non-symmetry and anti-packing pattern representation model(NAM).By describing the NAM and the method of the binary-bit plane decomposition(BPD)for the color image,a novel NAM-based representation algorithm for the color image is proposed.Also,the total data amount of the algorithm is analyzed.The theoretical and experimental results presented in this paper show that the NAM-based representation method can reduce the data storage much more effectively than the hierarchical structural linear quadtree-based representation method and is a better method to represent the color image pattern.The method is a new research area with respect to the color pattern representation and is valuable for the theoretical research and potential practical values such as decreasing the storage room,increasing the transmission speed,quickening the process procedure,matching pattern,and so on.

Abstract: This invention has provided a kind of processing method and system for the image registration of the synthetic aperture radar with high resolution. The method includes the following steps: (a) To build pyramid image and extract the mixture characteristics of the image, and then get the edge information and take it as matching input data. (b) To match the entire image rudely basing on the characteristics extracted. (c) To extract the feature corner of both the major and minor images after the rude match, and carry on the cross search for homonymic point to get the homonymic point pairs. (d) To divide the adaptive region according to the homonym point pairs. The system provided includes: the extraction cell for image data, the preprocessing cell for image and the matching cell for image. The matching precision at global sub-pixel level can be perfectly realized by this method and system. And so, it provide the base technology sustains for the further generalization of SAR data in the application fields such as map drawing, change detection, and the rebuilding of the three dimensional terrain.

Abstract: A medical image intensifying method of multi-dimension adaptive contrast transform includes decomposing medical image to be image set being set with resolution to be decreased gradually and being arranged in pyramid from, regulating the layering coefficient obtained from decomposition to intensify contrast of each layer image and local region at detailed gradation image, resynthesizing each detailed gradation image with coefficient being regulated to be image of original image being intensified.

Abstract: Accurate quality assessment of fused images, such as combined visible and infrared radiation images, has become increasingly important with the rise in the use of image fusion systems. We bring together three approaches, applying two objective tasks (local target analysis and global target location) to two scenarios, together with subjective quality ratings and three computational metrics. Contrast pyramid, shift-invariant discrete wavelet transform, and dual-tree complex wavelet transform fusion are applied, as well as levels of JPEG2000 compression. The differing tasks are shown to be more or less appropriate for differentiating among fusion methods, and future directions pertaining to the creation of task-specific metrics are explored.

Abstract: The implementation of several pyramid methods on programmable graphics processing units (GPUs) in recent years led to additional research interest in pyramid algorithms for real-time computer graphics. Of particular interest are efficient analysis and synthesis filters based on hardware-supported bilinear texture interpolation because they may be used as building blocks for many GPU-based pyramid methods. In this work, several new and extremely efficient GPU-implementations of pyramid filters are presented for the first time. The discussion employs a new notation, which was developed for the consistent and precise specification of these filters and also allowed us to systematically explore appropriate filter designs. The presented filters cover analysis and synthesis filters, (quasi-)interpolation and approximation, as well as discontinuous, continuous, and smooth filters. Thus, a toolbox of filters and their efficient implementations for a great variety of GPU-based pyramid methods is presented.

Abstract: Reliable feature extraction is crucial for accurate biometric recognition. Unfortunately feature extraction is hampered by noisy input data, especially so in case of fingerprints. We propose a method to enhance the quality of a given fingerprint with the purpose to improve the recognition performance. A Laplacian like image-scale pyramid is used for this purpose to decompose the original fingerprint into 3 smaller images corresponding to different frequency bands. In a further step, contextual filtering is performed using these pyramid levels and 1D Gaussians, where the corresponding filtering directions are derived from the frequency-adapted structure tensor. All image processing is done in the spatial domain, avoiding block artifacts while conserving the biometric signal well. We report on comparative results and present quantitative improvements, by applying the standardized NIST FIS2 fingerprint matcher to the FVC2004 fingerprint database along with our as well as two other enhancements. The study confirms that the suggested enhancement robustifies feature detection, e.g. minutiae, which in turn improves the recognition (20% relative improvement in equal error rate on DB3 of FVC2004).

Abstract: This paper deals with different techniques for registration and fusion of remote sensed images. In this work the lower spatial resolution multispectral and higher resolution panchromatic images of SPOT satellite are used. These images are registered using a registration algorithm that combines a simple yet powerful search strategy based on stochastic gradient with the similarity measure as mutual information, together with a wavelet-based multi-resolution pyramid. The algorithm is found to give sub pixel registration accuracy. The study is limited to pairs of images, which are misaligned by rotation and/or translation. The registered images are subjected to a pixel level multispectral image fusion process using wavelet transform approach. Spectral quality assessments shows that compared to other conventional image fusion techniques, this fusion process using wavelet transform keeps much of the spectral information in the merged image with respect to the original multispectral one. Finally, segmentation is performed on the fused images to validate the algorithms used for registration and fusion and the results show better accuracy for wavelet based methods than the conventional methods.

Abstract: A scanning device includes a scanning mechanism and a processing mechanism. The scanning mechanism scans an image fixed on a medium to generate a digital infrared representation of the image and a digital visible light representation of the image. The processing mechanism substantially reduces effects of noise and distortions within the digital visible light representation of the image in one pass. The processing mechanism at least decorrelates visible light aspects from the infrared representation of the image and employs a one-pass filter that uses both the infrared and the visible light representations of the image.

Abstract: A widely used hierarchical representation in many areas of computer vision and pattern recognition is the (regular) image pyramid, which employs both coarse to fine and fine to coarse processing strategies. Regular pyramids rapidly compute global information in a recursive manner, because their height is logarithmically bounded by the size of the input. Regular image pyramids lack shift invariance as a result of the fixed inter-level neighborhood. Irregular hierarchical structures (irregular pyramids) overcome shift invariance, among others. However, their logarithmic height cannot be guaranteed in general, as well as the computational efficiency. Main topics of this work are irregular graph pyramids and their application in image partitioning.We introduce two new decimationconcepts, maximal independent edge set (MIES) and maximal independent directed edge set (MIDES), both based on the maximal independent set principle.We show that the construction of stochastic irregular pyramids bounds logarithmically the height of the pyramid.Within this irregular graph pyramid framework, we introduce a time efficient image partitioning method based on the minimum spanning tree principle.