Chair: Constantine Kotropoulos, Aristotle University of Thessaloniki, Greece
Andrea Kutics, Japan Systems Co., Ltd. (Japan)
In this paper, an evolutionary method utilizing gray-scale morphological structures is proposed for object shape extraction. Artificial individuals are built up from gray-scale operator sequences by randomly choosing structuring elements from a basic operator pool. These individuals are mapped to list-like data representation structures and are manipulated by recombining or changing parts of the operator sequences randomly chosen. The extracted objects are obtained by carrying out the filtering with the best fit individual and calculating a residue image. The normalized correlation of the filtering results and the contributed input image areas is calculated for fitness. This method requires no preliminary knowledge of the object shape and also no constraints are used for image background and contrast. The evolutionary approach provides global and directed search on large number of possible morphological sequences and a method that can be applied on a wide range of images. The morphological operations are implemented by low level image processing steps such as image shifting and min-max operations and executed as parallel tasks on a multiprocessor basis by applying both data and algorithmic parallelization. As an example, this method is applied to the shape extraction of blood vessels in a system consisting of a camera device connected to a grid of transputer nodes.
Constantine Kotropoulos, Aristotle University of Thessaloniki (Greece)
Ioannis Pitas, Aristotle University of Thessaloniki (Greece)
A novel dynamic link architecture based on multiscale morphological dilation-erosion is proposed for face verification in a cooperative scenario where the candidates claim an identity that is to be checked. More specifically, a sparse grid is placed over the facial region of the image of each person in the reference set. Multiscale morphological operations are employed then to yield a feature vector at each grid node. Subsequently, dynamic link matching is applied to establish an isomorphism between the reference grid of the claimed person and the variable graph built over the face image of the candidate. The performance of the morphological dynamic link Architecture (MDLA) is evaluated in terms of the receiver operating characteristic (ROC) for several threshold selections on the matching error in the M2VTS database. The experimental results indicate that the proposed method outperforms the dynamic link matching with Gabor based feature vectors.
K. Sivakumar, The Johns Hopkins University (U.S.A.)
J. Goutsias, The Johns Hopkins University (U.S.A.)
Morphological size densities axe frequently used as descriptors of granularity or texture within an image. They have been successfully applied in many image processing and analysis tasks. It is extremely difficult however to analytically calculate the size density. In a previous work, we studied the problem of estimating the (discrete) morphological size density of random images, by means of empirical as well as Monte Carlo estimators. In this paper, we present applications of size density estimators to the problems of texture classification and morphological filtering. Experimental results demonstrate that texture classification, by means of morphological size densities, produces highly accurate results, even when the texture classes are chosen to be visually similar. For morphological filtering, we have obtained, by means of morphological size densities, two useful classes of filters, namely the class of generalized alternating filters and the class of generalized alternating sequential filters, which generalize the well known alternating filters and alternating sequential filters, respectively.
Lionel Merlat, French-German Research Institute of Saint-Louis (France)
Nicolas Silvestre, French-German Research Institute of Saint-Louis (France)
Jean Merckle, Universite de Haute-Alsace (France)
An image erosion architecture based on the CNN framework well suited for a VLSI implementation is presented. The erosion process can extract the center of gravity of an object coded by binary levels with a dynamical border peeling process. We first demonstrate that such an operator can not be implemented with the original model of CNN. Hence, the modification of the coupling function between cells is discussed. It is shown that an exclusive-or like gate (XOR) can be used to achieve a contour peeling process. The behavior of the new CNN model is investigated on a 1D chain of cells. The principle is then extended to a 2D map and is illustrated with a few numerical simulations. A reliable implementation approach of this CNN with its nonlinear cloning template is discussed in the appendix.
Nikos Nikopoulos, University of Thessaloniki (Greece)
Ioannis Pitas, University of Thessaloniki (Greece)
This paper proposes a fast algorithm for implementing the basic operation of Minkowski addition for the special case of binary three-dimensional images, using three-dimensional structuring elements of arbitrary size and shape. The application of the proposed algorithm for all the other morphological transformations is straightforward, as they can all be expressed in terms of Minkowski addition. The efficiency of the algorithm is analysed and some experimental results of its application are presented. As shown, the efficiency of the algorithm increases with the size of the structuring element.