Abstract: We address the problem of computing a general-purpose early visual representation that satisfies two criteria. 1) Explicitness: To be more useful than the original pixel array, the representation must take a significant step toward making important image structure explicit. 2) Completeness: To support a diverse set of high-level tasks, the representation must not discard information of potential perceptual relevance. The most prevalent representation in image processing and computer vision that satisfies the completeness criterion is the wavelet code. In this paper, we propose a very different code which represents the location of each edge and the magnitude and blur scale of the underlying intensity change. By making edge structure explicit, we argue that this representation better satisfies the first criterion than do wavelet codes. To address the second criterion, we study the question of how much visual information is lost in the representation. We report a novel method for inverting the edge code to reconstruct a perceptually accurate estimate of the original image, and thus demonstrate that the proposed representation embodies virtually all of the perceptually relevant information contained in a natural image. This result bears on recent claims that edge representations do not contain all of the information needed for higher level tasks.

Abstract: An original image is sharpened by obtaining a first frequency-domain representation of the original image, selecting one or more elements from this first representation based on one more criteria such as element magnitude and frequency, scaling the selected elements according to one or more scale factors, and forming a second frequency-domain representation by combining the scaled selected elements with the unselected elements of the first representation. A sharpened reproduction of the original image may be generated by applying an inverse transform to the second frequency-domain representation. A technique for deriving the value of the one or more scale factors is also discussed.

Abstract: A signal processing apparatus and concomitant method for learning and integrating features from multiple resolutions for detecting and/or classifying objects. The signal processing apparatus comprises a hierarchical pyramid of neural networks (HPNN) having a “fine-to-coarse” structure or a combination of the “fine-to-coarse” and the “coarse-to-fine” structures.

Abstract: We consider sparse image decomposition, in the hope that a sparser decomposition of an image may lead to a more efficient method of image coding or compression. Recently, many transforms have been proposed. Typically, each of them is good at processing one class of features in an image but not other features. For example, the 2-D wavelet transform is good at processing point singularities and patches in an image but not linear singularities, while a recently proposed method—the edgelet-like transform—is good for linear singularities, but not for points. Intuitively, a combined scheme may lead to a sparser decomposition than a scheme using only a single transform. Combining several transforms, we get an overcomplete system or dictionary. For a given image, there are infinitely many ways to decompose it. How to find the one with the sparsest coefficients? We follow the idea of Basis Pursuit—finding a minimum 1 norm solution. Some intuitive discussion and theoretical results show that this method is optimal in many cases. A big challenge in solving a minimum 1 norm problem is the computational complexity. In many cases, due to the intrinsic nature of the high-dimensionality of images, finding the minimum 1 norm solution is nearly impossible. We take advantage of the recent advances in convex optimization and iterative methods. Our approach is mainly based on two facts: first, we have fast algorithms for each transform; second, we have efficient iterative algorithms to solve for the Newton direction. The numerical results (to some extent) verify our intuitions, in the sense that: [1] the combined scheme does give sparser representations than a scheme applying only a single transform; [2] each transform in the combined scheme captures the features that this transform is good at processing. (Actually, [2] is an extension of [1].) With improved efficiency in numerical algorithms, this approach has the promise of producing more compact image coding and compression schemes than existing ones.

Abstract: Working with contrast instead of luminance can facilitate numerous image processing and analysis tasks. Unfortunately, a common definition of contrast suitable for all situations does not exist. We review existing contrast definitions for natural images and propose a new isotropic contrast measure, which is computed from oriented filters. We investigate some of its properties and apply it to natural images.

Abstract: Goal of this paper is to provide a quantitative performance evaluation of multiresolution schemes capable to carry out feature-based fusion of data collected by multispectral and panchromatic imaging sensors having different spectral and ground resolutions. To this aim a set of quantitative parameters has been recently proposed. Both visual quality, regarded as contrast, presence of fine details, and absence of impairments and artifacts (e.g., blur, ringing), and spectral fidelity (i.e., preservation of spectral signatures) are concerned and embodied in the measurements. Out of the three methods compared, respectively based on highpass filtering (HPF), wavelet transform (WT), and generalized Laplacian pyramid (GLP), the latter two are far more efficient than the former, thus establishing the advantages for data fusion of a formally multiresolution analysis.

Abstract: A method and apparatus for measuring stereo image disparity, for use in a 3-D modelling system. The method includes processing left and right camera images to form an image pyramid, calculating a disparity map at the coarsest level in the pyramid, and using this disparity map to carry out a warping operation on one of the images at the next-coarsest level, prior to calculating a disparity map for that level. This process is repeated for each subsequent pyramid level, at each level using the disparity map obtained at the previous level for carrying out the warping process, until a final disparity map for the least coarse pair of images in the pyramid is obtained. A computer program product for implementing this method is claimed, as well as a new method and apparatus for calibrating the cameras.

Abstract: akhencha I hbriand} @ireste.fr In the domain of the content-based image retrieval, the user formulates his queries from both visual and textual descriptions. In the sequel, we will only dwell on one of the most important visual features, namely the shape feature. The shape feature is essential as it corresponds to region of interest in images. Consequently, the shape representation is fundamental. This description must be compact and accurate, and it must own properties of invariance to several geometric transformations. After presenting several shape representations, we present the two complementary methods implemented in our prototype. The first one is an existing well-known approach, Freeman code, and the second one is an adaptation of a famous approach, Fourier theory. Simulations allow us to compare our results with results obtained under MATLAB, a powerful mathematical software, and to validate the proposed method.

Abstract: A low entropy pyramidal image data structure suited for lossless coding and progressive transmission is proposed in this work. The new coder, called the progressively predictive pyramid (PPP) is based on the well-known Laplacian pyramid. By introducing inter-resolution predictors into the original Laplacian pyramid, we show that the entropy level in the original pyramid can be reduced significantly. To take full advantage of progressive transmission, a scheme is introduced to create the predictor adaptively, thus eliminating the need to transmit the predictor and reducing the coding overheads. A method for designing the predictor is presented. Numerical results show that PPP is superior to traditional approaches to pyramid generation in the sense that the pyramids generated by PPP always have significantly lower entropy values.

Abstract: We consider a scale discrete wavelet approach on the sphere based on spherical radial basis functions. If the generators of the wavelets have a compact support, the scale and detail spaces are finite-dimensional, so that the detail information of a function is determined by only finitely many wavelet coefficients for each scale. We describe a pyramid scheme for the recursive determination of the wavelet coefficients from level to level, starting from an initial approximation of a given function. Basic tools are integration formulas which are exact for functions up to a given polynomial degree and spherical convolutions.

Abstract: Eddy current testing is one of the most popular nondestructive testing methods; a multifrequency testing is often applied to cancel the unwanted signals to improve the signal-to-noise ratio. This is usually accomplished by combining the results obtained at different frequencies in the spatial domain. In this paper a fusion strategy to integrate two-dimensional, multifrequency signals is introduced, which is based on the multiresolution analysis method. A signal-to-noise ratio criteria is adopted to evaluate the fusion results. Finally, the effectiveness of the pyramid method is discussed.

Abstract: Matching of a reference template with an image is a computationally expensive job. Particularly in fast real-time applications, large images and search ranges led to serious implementation problems. Therefore a reduction of the template size achieved by the selection of an appropriate subtemplate which is used for point correlation (subtemplate matching) may significantly decrease computational cost. In this paper a modified algorithm of the subtemplate point selection is proposed and explored. With the additional use of image pyramids, we can reduce the computational costs even further. The algorithm starts with a coarse search grid in the top level of the image pyramid generated for the full intended resolution. The procedure continues until the lowest level of the pyramid, the original image, is reached. The computational costs of this algorithm part satisfy the requirement for on- line processing. The preparation of the subtemplate for the point correlation is carried out in off-line mode, i.e., there is no rigorous limit of computational costs. The technique that applies the point correlation to image template matching within the image pyramid concept is proposed and the results obtained are discussed. It is especially useful for fast real- time system implementation when a large number of template matchings are needed in the same image.

Abstract: The neural abstraction pyramid is a hierarchical neural architecture for image interpretation that is inspired by the principles of information processing found in the visual cortex. In this paper we present an unsupervised learning algorithm for its connectivity based on Hebbian weight updates and competition. The algorithm yields a sequence of feature detectors that produce increasingly abstract representations of the image content. These representations are distributed and sparse, and facilitate the interpretation of the image. We apply the algorithm to a dataset of handwritten digits, starting from local contrast detectors. The emerging feature detectors correspond to step edges, lines, strokes, curves, and digit shapes. They can be used to reliably classify the digits.

Abstract: Local changes or variations of the intensity of an image (such as edges and corners), are important information for image processing and pattern recognition. Wavelet analysis is one of the most popular techniques that can be used to detect local intensity variation. In this paper, multiresolution analysis of wavelets is used to decompose images into pyramid images. Edges and peaks are extracted from pyramid images. A coarse-to-fine image segmentation method is proposed in this paper.

Abstract: This paper describes a document image analysis system using multiple agents working on a pyramid structure to separate text from graphics in the image Text strings appear as different groupings of connected components at different image resolutions As such, the pyramid structure, which is a multi-resolution image representation, provides a natural means of identifying and grouping of character strings in the document at different levels of resolution The pyramid structure is also amenable to parallel processing, where multiple agents in the system can individually and concurrently look for groups of connected components at appropriate levels The agent-based pyramid operations do not require expensive feature analysis among different connected components to detect text strings as found in other existing works

Abstract: Technological advances provide now the opportunity to automate the pavement distress assessment. This paper deals with an approach for achieving an automatic vision system for road surface classification. Road surfaces are composed of aggregates, which have a particular grain size distribution and a mortar matrix. From various physical properties and visual aspects, four road families are generated. We present here a tool using a pyramidal process with the assumption that regions or objects in an image rise up because of their uniform texture. Note that the aim is not to compute another statistical parameter but to include usual criteria in our method. In fact, the road surface classification uses a multiresolution cooccurrence matrix and a hierarchical process through an original intensity pyramid, where a father pixel takes the minimum gray level value of its directly linked children pixels. More precisely, only matrix diagonal is taken into account and analyzed along the pyramidal structure, which allows the classification to be made.

Abstract: We propose a face recognition system based on a biologically inspired filtering method. Our work differs from previous proposals in: 1) the multistage filtering method employed; 2) the pyramid structure used, and most importantly; 3) the prototype construction scheme to determine the models stored in memory. The method is much simpler than previous proposals and relatively inexpensive computationally, while attaining error rates as low as 5%, very close to the best reported results.

Abstract: This paper describes a pyramid-based method for clustering single polarization synthetic aperture radar (SAR) images. A pyramid of the image is first constructed by using the wavelet transform. The fuzzy c-means (FCM) algorithm is then applied to the pyramid with the coarse-to-fine approach. Finally, clustering is followed by a majority filtering to obtain more homogeneous regions. The algorithm has been tested on JERS-1/SAR images, and the results demonstrate its potential usefulness.

Abstract: Real-time image interpretation is required in many applications. Due to the computational complexity of the problem, dedicated architectures are needed for achieving real-time processing. In real-time applications, image frames are continuously fed to the computer vision system and three different levels of operations are performed on each image frame. Multi-layer systems, such as the pyramid architecture, provide an efficient means to solve those problems by exploiting parallelism in both spatial and temporal domain. We present techniques applied to solve an image processing problem using a three-layer architecture and demonstrate the system's ability to conduct these operations in real-time.

Abstract: Until recently, collections of digital images were stored in classical databases and indexed by keywords entered by a human operator. This is not longer practical, due to the growing size of these collections. Moreover, the keywords associated with an image are either selected from a fixed set of words and thus cannot cover the content of all images; or they are the operators' personal description of each image and, therefore, are subjective. That is why systems for image indexing based on their content are needed. In this context, we propose in this paper a new system, MUVIS*, for content-based indexing and retrieval for image database management systems. MUVIS*indexes by key words, and also allows indexing of objects and images based on color, texture, shape and objects' layout inside them. Due to the use of large vector features, we adopted the pyramid trees are used for creating the index structure. The block diagram of the system is presented and the functionality of each block is explained. The features used are presented as well.

Abstract: We formulate a model for probability distributions on image spaces. We show that any distribution of images can be factored exactly into conditional distributions of feature vectors at one resolution (pyramid level) conditioned on the image information at lower resolutions. We would like to factor this over positions in the pyramid levels to make it tractable, but such factoring may miss long-range dependencies. To fix this, we introduce hidden class labels at each pixel in the pyramid. The result is a hierarchical mixture of conditional probabilities, similar to a hidden Markov model on a tree. The model parameters can be found with maximum likelihood estimation using the EM algorithm. We have obtained encouraging preliminary results on the problems of detecting various objects in SAR images and target recognition in optical aerial images.

Abstract: Many autonomous vehicle navigation systems have adopted area-based stereo image processing techniques that use correlation measures to construct disparity maps as a basic obstacle detection and avoidance mechanism. Although the intra-scale area-based techniques perform well in pyramid processing frameworks, significant performance enhancement and reliability improvement may be achievable using wavelet- based inter-scale correlation measures. This paper presents a novel framework, which can be facilitated in unmanned ground vehicles, to recover 3D depth information (disparity maps) from binocular stereo images. We propose a wavelet- based coarse-to-fine incremental scheme to build up refined disparity maps from coarse ones, and demonstrate that usable disparity maps can be generated from sparse (compressed) wavelet coefficients. Our approach is motivated by a biological mechanism of the human visual system where multiresolution is known feature for perceptional visual processing. Among traditional multiresolution approaches, wavelet analysis provides a mathematically coherent and precise definition to the concept of multiresolution. The variation of resolution enables the transform to identify image signatures of objects in scale space. We use these signatures embedded in the wavelet transform domain to construct more detailed disparity maps at finer levels. Inter-scale correlation measures within the framework are used to identify the signature at the next finer level, since wavelet coefficients contain well-characterized evolutionary information.

Abstract: A system using multiple agents working on a pyramid structure to do text extraction is described in this paper. The method is based on the observation that text strings appear as different groupings of connected components at appropriate resolutions. The pyramid structure, which is a multi-resolution image representation, is amenable to parallel processing for detection of text strings. Agents in the system individually and concurrently look for groups of connected components at appropriate levels. They may in turn spawn new agents when connected components become disjointed at finer resolution levels. The agent-based pyramidal operations do not require expensive feature analysis among different connected components to detect text strings as found in other existing works.

Abstract: It is widely accepted that responses from on- and off-center cells give rise to edges and are equivalent to edge detectors. In this paper, we point out that on- and off-center cell responses provide more information than edges. We show that an edge-based representation makes the ownership of boundaries ambiguous and requires a combinatorial search to model perceptual grouping. By analyzing the differences between edges and responses from on- and off-center cells, we propose a boundary-pair representation, which makes the ownership of boundaries explicit and eliminates the need of a combinatorial search computationally. Each boundary in the boundary-pair representation is associated with regional attributes. We show that this representation is equivalent to a surface representation through a local diffusion. This provides a unified representation for perception modeling. Based on this representation, a figure-ground segregation network is constructed to demonstrate the capabilities of the model in explaining many perceptual phenomena.

Abstract: We present a Bayesian segmentation algorithm which is part of a fully Bayesian approach for automatic information extraction from satellite images. It was shown that pyramidal image models based on multi-scale Markov random fields in combination with a texture model yield good classification and segmenta­ tion results. The texture model is used for an initial characterization and then an optimal segmentation is inferred using the multi-scale random field defined on a pyramid structure. Segment probabilities are calculated in a fine-to-rough analysis and segmentation is performed by a rough-to-fine decision algorithm that maxi­ mizes the a posteriori probability for the pyramid. The procedure is iterated until it converges to a stable solution. We improve the maximization procedure by optimizing the underlying pyrami­ dal structure of the multi-scale Markov random field. Neighborhood dependencies are switched on and off according to the image data.