Search

Thinning is an iterative layer by layer erosion for extracting skeletons. This paper presents an efficient 3D parallel thinning algorithm which produces medial surfaces. A 3-subiteration strategy is proposed: the thinning operation is changed from iteration to iteration with a period of 3 according to the three deletion directions. [Copyright &y& Elsevier]

This paper reports the application of a recently developed shape-from-shading technique to estimate facial pose. The shape-from-shading algorithm uses a new geometric technique for solving the image irradiance equation together with curvature consistency constraints. Orientation histograms extracted from the the needle-maps delivered by the new shape-from-shading algorithm are used to estimate facial pose. We present a simple model of how the histogram bin-contents transform under rotation of the head. The parameters of this model are the head pose angles. We estimate pose by searching for the rotation angles which maximise the correlation between transformed histograms. A sensitivity analysis reveals that the methods can deliver pose estimates that are accurate within a few degrees. [Copyright &y& Elsevier]

This paper presents an automatic system for fire detection in video sequences. There are several previous methods to detect fire, however, all except two use spectroscopy or particle sensors. The two that use visual information suffer from the inability to cope with a moving camera or a moving scene. One of these is not able to work on general data, such as movie sequences. The other is too simplistic and unrestrictive in determining what is considered fire; so that it can be used reliably only in aircraft dry bays. We propose a system that uses color and motion information computed from video sequences to locate fire. This is done by first using an approach that is based upon creating a Gaussian-smoothed color histogram to detect the fire-colored pixels, and then using a temporal variation of pixels to determine which of these pixels are actually fire pixels. Next, some spurious fire pixels are automatically removed using an erode operation, and some missing fire pixels are found using region growing method. Unlike the two previous vision-based methods for fire detection, our method is applicable to more areas because of its insensitivity to camera motion. Two specific applications not possible with previous algorithms are the recognition of fire in the presence of global camera motion or scene motion and the recognition of fire in movies for possible use in an automatic rating system. We show that our method works in a variety of conditions, and that it can automatically determine when it has insufficient information. [Copyright &y& Elsevier]

It is of utmost importance in multimedia processing to achieve still image segmentation, i.e., to partition images into regions of coherent color and texture. In this paper we propose a novel image segmentation method using a special visual descriptor. For each pixel p, the discrete cosine transform (DCT) of the block centered on p together with its location in the image is employed as its content descriptor thus resulting in a long vector v→p, referred to as situational DCT descriptors (SDDs). A scalar quantization step is then carried out on the DCT component of SDDs to reflect the fact that the human vision system is not of uniform discriminative sensitivity to details of different frequencies. Next the principal component analysis is conducted to drastically reduce the dimensionality of SDDs. The adaptive K-means algorithm is then performed to arrive at the region assignment for each pixel. The final partitioning results are obtained after performing the post-processing step. Encouraging empirical performance has been demonstrated. [Copyright &y& Elsevier]

Recently, a new image registration method, based on the Hausdorff fraction and a multi-resolution search of the transformation space, has been developed in the literature. This method has been applied to problems involving translations, translation and scale, and affine transformations. In this paper, we adapt the above method to the set of similarity transformations. We also introduce a new variant of the Hausdorff fraction similarity measure based on a multi-class approach, which we call the multi-class Hausdorff fraction (MCHF). The multi-class approach is more efficient because it matches feature points only if they are from the same class. To validate our approach, we segment edge maps into two classes which are the class of straight lines and the class of curves, and we apply the new multi-class approach to two image registration examples, using synthetic and real images, respectively. Experimental results show that the multi-class approach speeds up the multi-resolution search algorithm. [Copyright &y& Elsevier]

The knowledge base of a probabilistic expert system is usually represented as a Bayesian network. Most of the knowledge engineering tools used in the development of probabilistic expert systems do not carry out the inference process directly over the network, but in a secondary graphical structure called a junction tree. The efficiency of inference (propagation) algorithms depends on the size of the junction tree obtained, and this size depends on the elimination sequence used during the compilation/transformation of the Bayesian network into a junction tree. The problem of searching for the best elimination sequence is an NP-hard problem [W. Wen, in: P. Bonissone, M. Henrion, L. Kanal and Z. Lemmer (Eds.), Uncertainty in Artificial Intelligence, vol. 6, North-Holland, Amsterdam, 1991, pp. 209–224], and this has motivated the proliferation of approximate methods to approach it (based variously on greedy heuristics, genetic algorithms, simulated annealing, etc.). In this paper we investigate the applicability to this problem of a new combinatorial optimization technique, inspired by a natural model, which has appeared recently: ant colony optimization [M. Dorigo, Optimization, learning and natural algorithms, Ph.D. thesis, Politecnico di Milano, Italy, 1992; M. Dorigo and L. Gambardella, IEEE Trans. Evol. Comput. 1 (1997) 53; M. Dorigo and G.Di. Caru, in: New Ideas in Optimization, McGraw-Hill, New York, 1999]. Our approach is validated by using a set of complex networks obtained from a repository. [Copyright &y& Elsevier]

In this paper, we propose a wavelet decomposition approach for rotation-invariant template matching. In the matching process, we first decompose an input image into different multi-resolution levels in the wavelet-transformed domain, and use only the pixels with high wavelet coefficients in the decomposed detail subimage at a lower resolution level to compute the normalized correlation between two compared patterns. To make the matching invariant to rotation, we further use the ring-projection transform, which is invariant to object orientation, to represent an object pattern in the detail subimage. The proposed method significantly reduces the computational burden of the traditional pixel-by-pixel matching. Experimental results on a variety of real images have shown the efficacy of the proposed method. [Copyright &y& Elsevier]

This paper presents a new fast front propagation algorithm for image segmentation. To approximate the partial differential equation (PDE) in level set algorithm, instead of moving the front in a small constant time step, the point with a minimum arrival time will be touched in one iteration. Only in a neighbourhood of this point, should the level set function be updated. Like the previously proposed level set methods, it is a robust method for image segmentation with capabilities to handle topological changes, significant protrusions and narrow regions. It is faster than the narrow band algorithm and more robust than the monotonically advancing scheme in image segmentation. The effectiveness and the capabilities of the algorithm were verified by simulated and real experiments. [Copyright &y& Elsevier]