International Journal of Pattern Recognition and Artificial Intelligence

Correspondence is one of the major problems that must be solved in stereo vision. Correlation has been commonly used in the past for this problem. However, most classical linear correlation methods fail near depth discontinuities and in the presence of occlusions. Many robust methods have been proposed that claim to effectively deal with some or all of these issues. Many of these robust methods are transformation-based, however, other robust methods are non-transformation based. This paper gives five requirements that should be met by a transformation-based robust correlation method. We compare some of the robust correspondence methods and demonstrate their utility on different data sets. Based on these results, we propose a solution to the correspondence problem which represents a compromise between the speed of classical correlation and the improved results obtained from a more robust correspondence method. Also, we propose a median filtering technique that removes noise from the disparity maps while preserving certain image features usually removed by ordinary median filtering.

In this paper, we propose a novel area-based algorithm for stereo correspondence estimation. The new dense correspondence estimation algorithm, which we call HACM (Hierarchical Adaptive Curve Matching), is suitable for stereo analysis of both static and dynamic scenes. The core of this method is the derivation of a pixel-based dense adaptive curve representation. For each pixel in the concerned 2-D images, the changes are characterized roughly in the local surface shape, the texture direction, and the luminance properties, all at a slight overhead cost. The matching is carried out in a hierarchical manner that incorporates several novel ideas that improve its efficiency and robustness. Experiments conducted on both synthetic and natural stereo pairs, and natural video streams, reveal the favorable performance of this new algorithm.

Feature point matching is a key step for most problems in computer vision. It is an ill-posed problem and suffers from combinatorial complexity which becomes even more critical with the increase in data and the presence of outliers. The work covered in this paper describes a new framework to solve this problem in order to achieve robust registration of two feature point sets assumed to be available. This framework combines the use of extremal optimization heuristic with a clever startup routine which exploits some properties of singular value decomposition. The role of the latter is to produce an interesting matching configuration whereas the role of the former is to refine the initial matching by generating hypothetical matches and outliers using a far-from-equilibrium based stochastic rule. Experiments on a wide range of real data have shown the effectiveness of the proposed method and its ability to achieve reliable feature point matching.

In this paper, we propose a new method to simultaneously achieve segmentation and dense matching in a pair of stereo images. In contrast to conventional methods that are based on similarity or correlation techniques, this method is based on geometry, and uses correlations only on a limited number of key points. Stemming from the observation that our environment is abundant in planes, this method focuses on segmentation and matching of planes in an observed scene. Neither prior knowledge about the scene nor camera calibration are needed. Using two uncalibrated images as inputs, the method starts with a rough identification of a potential plane, defined by three points only. Based on these three points, a plane homography is then calculated and, used for validation. Starting from a seed region defined by the original three points, the method grows the current region by successive move/confirmation steps until occlusions and/or surface discontinuity occur. In this case, the homography-based mapping of points between the two images will not be valid anymore. This condition is detected by the correlation, used in the confirmation process. In particular, this method grows a region even across different colors as long as the region is planar. Experiments on real images validated our method and showed its capability and performance.

In this paper we present a novel technique for non-rigid medical image registration and correspondence finding based on a multiple-layer flexible mesh template matching technique. A statistical anatomical model is built in the form of a tetrahedral mesh, which incorporates both shape and density properties of the anatomical structure. After the affine transformation and global deformation of the model are computed by optimizing an energy function, a multiple-layer flexible mesh template matching is applied to find the vertex correspondence and achieve local deformation. The multiple-layer structure of the template can be used to describe different scale of anatomical features; furthermore, the template matching is flexible which makes the correspondence finding robust. A leave-one-out validation has been conducted to demonstrate the effectiveness and accuracy of our method.

Image registration plays a crucial role in the computer vision and medical imaging field where it is used to develop a spatial mapping between different sets of data. These transformations can range from simple rigid registrations to complex nonrigid deformations. Mutual information (MI) is a popular entropy-based similarity measure which has recently experienced a prolific expansion in a number of image registration applications. Stemming from information theory, this measure generally outperforms most other intensity-based measures in multimodal applications as it only assumes a statistical dependence between images. This paper provides a thorough introduction to the MI measure and its use in rigid medical image registration. A look at the extensions proposed to the original measure will also be provided. These were developed to improve the robustness of the measure and to avoid certain cases when maximizing MI does not lead to the correct spatial alignment.

Computer support for early detection of breast cancer requires a proper mimicking of the way radiologists compare mammographic images; by comparing bilateral (images of the left and right breasts) and temporal images. In this paper, one method for bilateral registration and intensity normalization and two methods for difference analysis are described. The bilateral registration is based on anatomical features and assumptions of how the female breast is deformed under compression. The first method for differential analysis is based on the absolute difference between the registered images while the second method is based on statistical differences between properties of corresponding neighborhoods. The methods are tested on images from the MIAS database (on 100 images with 59 abnormalities distributed over four types) and evaluated by FROC-analysis. The performances of the two methods are similar but the statistical method gives better performance at a lower false positive rate and is better in particular for detecting asymmetrical developments.

This paper deals with the segmentation problem of cervical cell images. This segmentation consists of separating each cell into its core and its cytoplasm. The mentioned separation is part of a uterus cancer recognition system based on the morphology of both the cell core and the cell cytoplasm. In this paper, we propose to perform the above separation by using a multifractal algorithm. An important feature of the proposed algorithm is its low computation cost. To increase the quality of the segmentation, we propose an optimization step based on genetic algorithms. The proposed processing has been tested on several images. Herein, we present only the results obtained for three different cell images.

Model-Based 3D object recognition systems have a variety of potential applications, but widespread use of such systems has not occurred, due to a number of factors including the representational limitations of models. One historical limitation is the discriminatory representation of free-form objects. The system described in this paper recognizes free-form objects in dense range data acquired by a structured light rangefinder. Images and object models are represented as a network of salient segments which are then brought into correspondence until a reliable pose estimate is available. Experiments with a database of images and object models highlight the contributions of this system.

We describe an object recognition system based on symbolic contour graphs. The image to be analyzed is transformed into a graph with object corners as vertices and connecting contours as edges. Image corners are determined using a robust multiscale corner detector. Edges are constructed by line-following between corners based on evidence from the multiscale Gabor wavelet transform. Model matching is done by finding subgraph isomorphisms in the image graph. The complexity of the algorithm is reduced by labeling vertices and edges, whereby the choice of labels also makes the recognition system invariant under translation, rotation and scaling. We provide experimental evidence and theoretical arguments that the matching complexity is below O(#V³), and show that the system is competitive with other graph-based matching systems.