We present a novel statistical and variational approach to image segmentation based on a new algorithm named region competition. This algorithm is derived by minimizing a generalized Bayes/MDL criterion using the variational principle. The algorithm is guaranteed to converge to a local minimum and combines aspects of snakes/balloons and region growing. Indeed the classic snakes/balloons and region growing algorithms can be directly derived from our approach. We provide theoretical analysis of region competition including accuracy of boundary location, criteria for initial conditions, and the relationship to edge detection using filters. It is straightforward to generalize the algorithm to multiband segmentation and we demonstrate it on grey level images, color images and texture images. The novel color model allows us to eliminate intensity gradients and shadows, thereby obtaining segmentation based on the albedos of objects. It also helps detect highlight regions. 1 Division of Appli...

Discontinuity-preserving Bayesian image restoration typically involves two Markov random fields: one representing the image intensities/gray levels to be recovered and another one signaling discontinuities/edges to be preserved. The usual strategy is to perform joint maximum a posteriori (MAP) estimation of the image and its edges, which requires the specification of priors for both fields. In this paper, instead of taking an edge prior, we interpret discontinuities (in fact their locations) as deterministic unknown parameters of the compound Gauss--Markov random field (CGMRF), which is assumed to model the intensities. This strategy should allow inferring the discontinuity locations directly from the image with no further assumptions. However, an additional problem emerges: The number of parameters (edges) is unknown. To deal with it, we invoke the minimum description length (MDL) principle; according to MDL, the best edge configuration is the one that allows the shortest description of the image and its edges. Taking the other model parameters (noise and CGMRF variances) also as unknown, we propose a new unsupervised discontinuity-preserving image restoration criterion. Implementation is carried out by a continuation-type iterative algorithm which provides estimates of the number of discontinuities, their locations, the noise variance, the original image variance, and the original image itself (restored image). Experimental results with real and synthetic images are reported.

In this paper, we develop a higher-order statistical theory of matching models against images. The basic idea is not only to take into account how much of an object can be seen in the image, but also what parts of it are jointly present. We showthat this additional information can improve the speci city (i.e., reduce the probability of false positive matches) of a recognition algorithm. We demonstrate formally that most commonly used quality of match measures employed by recognition algorithms are based on an independence assumption. Using the Minimum Description Length (MDL) principle and a simple scene-description language as a guide, we show that this independence assumption is not satis ed for common scenes, and propose several important higher-order statistical properties of matches that approximate some aspects of these statistical dependencies. We haveimplemented a recognition system that takes advantage of this additional statistical information and demonstrate its e cacy in comparisons with a standard recognition system based on bounded error matching. We also observe that the existing use of grouping and segmentation methods has signi cant e ects on the performance of recognition systems that are similar to those resulting from the use of higher-order statistical information. Our analysis provides a statistical framework in which to understand the effects of grouping and segmentation on recognition and suggests ways to take better advantage of such information.

We present a novel statistical and variational approach to image segmentation based on a new algorithm named region competition. This algorithm is derived by minimizing a generalized Bayes/MDL criterion using the variational principle. The algorithm is guaranteed to converge to a local minimum and combines aspects of snakes/balloons and region growing. Indeed the classic snakes/balloons and region growing algorithms can be directly derived from our approach. We provide theoretical analysis of region competition including accuracy of boundary location, criteria for initial conditions, and the relationship to edge detection using filters. It is straightforward to generalize the algorithm to multiband segmentation and we demonstrate it on grey level images, color images and texture images. The novel color model allows us to eliminate intensity gradients and shadows, thereby obtaining segmentation based on the albedos of objects, and it also helps detect highlight regions. A short version ...

We present a method for restoration of noisy tomographic images for detecting thin objects, such as explosives. Use of a weighted mean-square estimate optimizes the solution to place emphasis on the infrequent, but significant local structure associated with thin objects. Experimental results show successful restoration at very high noise levels. 1 Introduction Detection of certain objects in multidimensional image data is a common problem in industrial inspection. The objects may be a normal part of the subject of the inspection, or they may be anomalies. Detection is more difficult if the object has been deliberately concealed, which is likely with explosives or illegal drugs. A common method used to conceal malleable material is to form it into a thin sheet. It is well known that this is effective for conventional X-ray imagery where a thin object is almost invisible unless viewed on end. The use of tomography offers a solution to the problem of imaging thin sheets by viewing the ...

Abstract We present a method for restoration of noisy tomo-graphic images for detecting thin objects, such as explosives. Use of a weighted mean-square estimateoptimizes the solution to place emphasis on the infrequent, but significant local structure associated withthin objects. Experimental results show successful restoration at very high noise levels.

Machine vision systems are often considered to be composed of two subsystems: low-level vision and high-level vision. Low-level vision consists primarily of image processing operations performed on the input image to produce another image with more favorable character-istics. These operations may yield images with reduced noise or cause certain features of the image to be empha-sized (such as edges). High-level vision includes object recognition and, at the highest level, scene interpretation. The bridge between these two subsystems is the segmen-tation system. Through segmentation, the enhanced input image is mapped into a description involving regions with common features which can be used by the higher level vision tasks. There is no theory on image segmentation. Instead, image segmentation techniques are basically ad hoc and differ mostly in the way they emphasize one or more of the desired properties of an ideal segmenter and in the way they balance and compromise one desired property against another. These techniques can be categorized in a number of different groups including local vs. global, parallel vs. sequential, contextual vs. non contextual, interactive vs. automatic. In this paper, we categorize the schemes into three main groups: pixel-based, edge-based, and region-based. Pixel-based segmentation schemes classify pixels based solely on their gray levels. Edge-based schemes first detect local discontinuities (edges) and then use that information to separate the image into regions. Finally, region-based schemes start with a seed pixel (or group of pixels) and then grow or split the seed until the original image is composed of only homogeneous regions. Because there are a number of survey papers available, we will not discuss all segmentation schemes. Rather than a survey, we take the approach of a detailed overview. We focus only on the more common approaches in order to give the reader a flavor for the variety of techniques available yet present enough details to facilitate imple-mentation and experimentation.