Home Home
Reiner Lenz, Linköping University (Sweden)
Peter Meer, Rutgers University (U.S.A.)
The information in a color image is always a function of the illuminating source, the geometry, the reflectance properties of the object and the characteristic of the camera. Separating the influence of the spectral distribution of the illumination and the reflectance properties of the object is known as the color constancy problem. Successful separation is important for vision and pattern recognition tasks, quality control in the graphic arts and image database applications. We describe an approach to the color constancy problem which is based on statistical assumptions about the distribution of colors. It uses the eigenvector system of the logarithmic spectra in a large database of color samples and employs methods from robust statistics to recover the illumination spectrum. We illustrate the performance of the algorithm with a simulation in which the effect of the illumination by the standard A-source is eliminated.
ic973141.pdf ic973141.pdf
TOP
Christian Scheering, University of Bielefeld (Germany)
Alois Knoll, University of Bielefeld (Germany)
We present an efficient method for segmenting colour-images, which may be utilised in several robotic vision tasks. It categorises pixels according to their perceptual colour by exploiting the chromaticity contained in the signal of a standard colour camera as an index into a pre-clustered chromaticity plane. A technique called perceptual colour grouping is introduced to prevent oversegmentation. Experimental data demonstrate the performance of the proposed approach; computation time is reduced by a factor of 8...30 over previously known methods.
ic973145.pdf ic973145.pdf TOP
Jong-Kae Fwu, State University of New York at Stony Brook (U.S.A.)
Petar M. Djuric, State University of New York at Stony Brook (U.S.A.)
In this paper cluster validation criteria for piecewise constant image segmentation are proposed. All the criteria are based on the maximum a posteriori (MAP) principle and derived and implemented by four different, but related approaches. They are obtained by using Taylor expansions and three of them are derived by Bayesian predictive densities. The third and fourth criteria are implemented by the bootstrap technique, and their evaluations are, therefore, computationally more intensive than the evaluations of the first two. The proposed rules are compared by computer simulations with the widely used AIC and MDL criteria.
ic973149.pdf ic973149.pdf TOP
Shigeki Doi, Nara National Colleage of Technology (Japan)
Etsuko Ueda, Nara National Colleage of Technology (Japan)
Shunsuke Doi, Nara National Colleage of Technology (Japan)
A directional morphological operation can be performed by utilizing a rotational structuring element with a directional information, and its associated processing methods for overlapping and enclosing are proposed. As an example of overlapping method derived from the directional morphological operation, a distribution of high/low atmospheric pressures is obtained from the wind direction of weather report. The enclosing method obtained from the directional morphological operation is applied to a shape recognition system utilizing multi-ultrasonic sensor, and also this method is applied a immunological image processing which utilizes the function of self or non-self discrimination in immune system of a living body. From simulation and experimental results, it has been cleared that these methods are effective for image to extract mutual relationship among data including directional information and that object categories are discriminated by applying directional morphological operation for immunological discrimination.
ic973153.pdf ic973153.pdf TOP
Robert Porter, University of Bristol (U.K.)
Nishan Canagarajah, University of Bristol (U.K.)
The importance of texture analysis and classification in image processing is well known. However, many existing texture classification schemes suffer from a number of drawbacks. A large number of features are commonly used to represent each texture and an excessively large image area is often required for the texture analysis, both leading to high computational complexity. Furthermore, most existing schemes are highly orientation dependent and thus cannot correctly classify textures after rotation. In this paper, two novel feature extraction techniques for rotation invariant texture classification are presented. These schemes, using the wavelet transform and Gaussian Markov random field modelling, are shown to give a consistently high performance for rotated textures in the presence of noise. Moreover, they use just four features to represent each texture and require only a 16x16 image area for their analysis leading to a significantly lower computational complexity than most existing schemes.
ic973157.pdf TOP
Philippe Bolon, Univ. de Savoie (France)
R. Kara Falah, Univ. de Savoie (France)
In this paper, we introduce a method aiming at improving the segmentation results obtained with complex natural images. It is based on the integration of a set of image segmentation maps. This technique allows the redundancy between some primary segmentations to be extracted. Our approach is mainly region-oriented. It is based on a fuzzy description of the segmentation by means of a mechanism associating regions of different maps. Introducing a dissimilarity measurement allows a so-called segmentation filtering to be performed. This approach turns out to be very efficient in the case of natural complex images. It can be regarded as an extension of nonlinear filtering techniques.
ic973161.pdf TOP
Rüdiger Beß, IMMD 5, Univ. Erlangen (Germany)
Dietrich Paulus, IMMD 5, Univ. Erlangen (Germany)
Michael Harbeck, IMMD 5, Univ. Erlangen (Germany)
In this paper we describe a segmentation approach improving the computation of depth from stereo images compared to straight line segmentation. Using a straight line -- circular arc approximation of chain coded lines the computational effort is reduced significantly as well as the frequency of erroneous matches. We describe the result of a parallel implementation using object--oriented programming techniques. In segmentation as well as in matching we evaluate color information to improve accuracy and reliability of the depth values. The algorithm explained in this paper is part of a system computing depth from monocular image sequences. Each two consecutive images are considered as a stereo image. The depth images computed from these stereo images are fused to one complete depth map of the object surface. The results show substantial improvements in comparison to a monochrome system with respect to speed, accuracy, and completeness.
ic973165.pdf ic973165.pdf TOP
Albert Oliveras, UPC-Barcelona (Spain)
Philippe Salembier, UPC-Barcelona (Spain)
Luis Garrido, UPC-Barcelona (Spain)
Connected operators are increasingly used in image processing due to their properties of simplifying the image with various criteria, without loosing contour's information. These properties are related to the connected operator approach that either preserves or completely eliminates a determined connected component, according to an established criterion of analysis. In this paper we will define a new connected operator for stereo images. The goal is to simplify one of the images (left) in the sense that the operator will eliminate the image components that are not present at a determined location in the other image (right). This filter let us select in a stereo image, objects as a function of their distance from the observer (for instance used in auto guided vehicles).
ic973169.pdf ic973169.pdf TOPJoachim Hornegger, IMMD 5, Univ. Erlangen (Germany)
A new Bayesian framework for 3--D object classification and localization is introduced. Objects are represented as probability density functions, and observed features are treated as random variables. These probability density functions turn out a non geometric nature of models and characterize the statistical behavior of local object features like points or lines. The parameterization of model densities covers several terms of object recognition: locations and instabilities of features, rotation and translation, projection, the assignment of image and model features, as well as relations. This paper treats especially the probabilistic modeling of relational dependencies between single features. The mathematical framework, the training algorithms, as well as the localization and classification modules are discussed in detail. The experimental evaluation shows the usefulness of the introduced concepts on real image data.
ic973173.pdf ic973173.pdf TOPAndrás Kornai, IBM Almaden (U.S.A.)
It is almost universally accepted in speech recognition that phone- or word-level segmentation prior to recognition is neither feasible nor desirable, and in the dynamic (pen-based) handwriting recognition domain the success of segmentation-free techniques points to the same conclusion. But in image-based handwriting recognition, this conclusion is far from being firmly established, and the results presented in this paper show that systems employing character-level presegmentation can be more effective, even within the same HMM paradigm, than systems relying on sliding window feature extraction. We describe two variants of a Hidden Markov system recognizing handwritten addresses on US mail, one with presegmentation and one without, and report results on the CEDAR data set.
ic973177.pdf ic973177.pdf TOP
Erhan Öztop, METU (Turkey)
Adem Y. Mülayim, METU (Turkey)
Volkan Atalay, METU (Turkey)
Fatoş Yarman-Vural, METU (Turkey)
This paper describes a new framework, called, Repulsive Attractive (RA) Network for Baseline Extraction on document images. The RA network is a self organizing feature detector which interacts with the document text image through the attractive and repulsive forces defined among the network components and the document image. Experimental results indicate that the network can successfully extract the baselines under heavy noise and with overlaps between the ascending and descending portions of the characters of adjacent lines. The proposed method is also applicable to a wide range of image processing applications, such as curve fitting, segmentation and thinning.
ic973181.pdf ic973181.pdf TOP
Alessandro L. Koerich, DECOM, FEE, UNICAMP (Brazil)
Luan L. Lee, DECOM, FEE, UNICAMP (Brazil)
In this paper a scheme for bank cheque images compression based on layout knowledge is proposed. The layout structure of the cheques is analyzed and the non-essential parts are located. These parts, said, the background and the printed information are eliminated from the original image. The resulting image contains some noise that are eliminated by a filtering operation. The image is enclosed to eliminate some no informative parts. The final image has only the filled information. The digitized image can be easily reconstructed by restoring the filled information and summing it with background and printed information. The proposed compression scheme is tested by Brazilian bank cheques. Comparisons with other compression schemes, shows that the proposed scheme performs significantly better in terms of the compression efficiency, maintaining the visual quality.
ic973185.pdf ic973185.pdf TOP