Hierarchical Modeling of Multidimensional Data in Regularly Decomposed Spaces: Synthesis and Perspective

This fourth and last tome is focusing on describing the envisioned works for a project that has been presented in the preceding tome. It is about a new approach dedicated to the coding of still and moving pictures, trying to bridge the MPEG-4 and MPEG-7 standard bodies. The aim of this project is to define the principles of self-descriptive video coding. In order to establish them, the document is composed in five chapters that describe the various envisioned techniques for developing such a new approach in visual coding: - image segmentation, - computation of visual descriptors, - computation of perceptual groupings, - building of visual dictionaries, - picture and video coding. Based on the techniques of multiresolution computing, it is proposed to develop an image segmentation made from piecewise regular components, to compute attributes on the frame and the rendering of so produced shapes, independently to the geometric transforms that can occur in the image plane, and to gather them into perceptual groupings so as to be able in performing recognition of partially hidden patterns. Due to vector quantization of shapes frame and rendering, it will appear that simple shapes may be compared to a visual alphabet and that complex shapes then become words written using this alphabet and be recorded into a dictionary. With the help of a nearest neighbour scanning applied on the picture shapes, the self-descriptive coding will then generate a sentence made from words written using the simple shape alphabet.

Hierarchical Modeling of Multidimensional Data in Regularly Decomposed Spaces: Main Principles

The described works have been carried out in the framework of a mid-term study initiated by the Centre Electronique de l'Armement and led by ADERSA, a French company of research under contract. The aim was to study the techniques of regular dividing of numerical data sets so as to provide tools for problem solving enabling to model multidimensional numerical objects and to be used in computer-aided design and manufacturing, in robotics, in image analysis and synthesis, in pattern recognition, in decision making, in cartography and numerical data base management. These tools are relying on the principle of regular hierarchical decomposition and led to the implementation of a multidimensional generalization of quaternary and octernary trees: the trees of order 2**k or 2**k-trees mapped in binary trees. This first tome, dedicated to the hierarchical modeling of multidimensional numerical data, describes the principles used for building, transforming, analyzing and recognizing patterns on which is relying the development of the associated algorithms. The whole so developed algorithms are detailed in pseudo-code at the end of this document. The present publication especially describes: - a building method adapted disordered and overcrowded data streams ; - its extension in inductive limits ; - the computation of the homographic transformation of a tree ; - the attribute calculus based on generalized moments and the provision of Eigen trees ; - perception procedures of objects without any covering in affine geometry ; - several supervised and unsupervised pattern recognition methods.

Hierarchical Modeling of Multidimensional Data in Regularly Decomposed Spaces: Applications in Image Analysis

This last document is showing the gradual introduction of hierarchical modeling techniques in image analysis. The first chapter is dealing with the first works carried out in the field of industrial applications of pattern recognition. The second chapter is focusing on the usage of these techniques in satellite imagery and on the development of a satellite data archiving system in the aim of using it in digital geography. The third chapter is about face recognition based on planar image analysis and about the recognition of partially hidden patterns. The present publication is ending with the description of a future system of self-descriptive coding of still or moving pictures in relation with the current video coding standards. As in the previous documents, it will be found in annex algorithms targeted on image analysis according two complementary approaches: - boundary-based approach for the industrial applications of artificial vision; - region-based approach for satellite image analysis.