Descriptive Image Analysis: Part IV. Information Structure for Generating Descriptive Algorithmic Schemes for Image Recognition

This article is the fourth in a series of publications devoted to the state of the art and prospects for developing Descriptive Image Analysis, one of the leading and intensively developing fields of modern mathematical image analysis theory. The fundamental problem discussed in the article is to automate information extraction from images necessary for making intelligent decisions. This study is devoted to regularizing the generation of descriptive algorithmic image analysis and recognition schemes. The main result is the definition of a new mathematical structure with the following functional capabilities: 1) solution of an image recognition problem in a given formulation, with given initial data and a scenario that determines the sequence of application of information processing procedures and their iterative loops; 2) construction of descriptive algorithmic schemes for solving a problem with given initial data in the absence of a given scenario; in the case of a successful solution, the fixation of the sequence of procedures and information processing loops that yielded its solution governs the corresponding descriptive algorithmic schemes and scenarios that can be further used to solve the corresponding class of image recognition problems; 3) comparative analysis and optimization of methods for solving image recognition problems via their realization as descriptive algorithmic schemes and scenarios allowed by the structure. The introduced structure is a tool for representing and implementing information processing while solving an image recognition problem for arbitrary formulations, scenarios, models, and solution methods; it can also emulate any descriptive algorithmic scheme and combinations thereof, which are used and generated when solving an image recognition problem. The introduced structure is interpreted as a fundamental model for generating and emulating image recognition procedures. A type characteristic of the introduced information structure for generating descriptive algorithmic schemes is as follows: 1) the set of structure elements consists of two subsets: a) a subset of functional blocks that perform mathematical operations of information processing necessary to implement the used of processing, analysis, and image recognition methods; b) a subset of control blocks for information processing procedures, which verify the logical conditions for branching of processing procedures, whether the rules for stopping information processing are met, etc.; 2) relations given over the elements of the set of the structure, mainly, the partial ordering relations that determine the sequence of execution and methods for combining the functional and control blocks of the structure; 3) these relations, by definition, must satisfy the axioms of the Descriptive Approach to Image Analysis and Understanding. The paper presents the basic definitions associated with the introduction of a new information structure and describes the information processing procedures implemented therein, as well as the main blocks and loops. Block diagrams of the information structure and iteration loops are given. The fundamental importance of the results of the described studies for developing the mathematical theory of image analysis and their scientific novelty are related to formulation of problems and development of methods for modeling the processes for automating image analysis when the input data are poorly formalized representations of images, including spatial data—images and their fragments, image models, incompletely formalized representations, and subsets of combinations thereof. Introduction of a new information structure as a standard structure for a) representation of algorithms for analysis and recognition of 2-dimensional information and representation of procedures for constructing dual representations of images as descriptive algorithmic schemes; b) generation of descriptive algorithmic schemes indicated in (a) will make it possible to generalize and substantiate the known heuristic recognition algorithms, to comparatively analysis them, and to assess their mathematical properties and applied usefulness. The problem of automating the construction and analysis of descriptive algorithmic image recognition schemes is formulated for the first time.