17 - Genetic Image Interpretation

Milan Sonka

doi:10.1016/B978-012443880-4/50061-2

Image interpretation is the process of understanding the meaning of an image by identifying significant objects in the image and analyzing their spatial relationships. The need for reliable automated image interpretation methods is quickly growing, and image interpretation remains an area of active research. Traditional methods of image interpretation generally include segmentation, object recognition, and image interpretation steps. Segmentation refers to the process of dividing an image into regions that correspond to individual objects or structures. Accurate segmentation is critical to the success of any image interpretation method. Image interpretation can best be done by incorporating information about individual objects and their relationships into the process. Image segmentation represents a crucial preliminary stage of image interpretation. Segmentation is a difficult problem and rarely results in unambiguous and complete segmentation. The goal of image interpretation is to assign a label to each image object. The labeling should be consistent with the available a priori knowledge and, should favor more probable interpretations if there are multiple options. In case images have been previously segmented into meaningful objects, two main approaches exist to achieve image interpretation: discrete and probabilistic labeling approaches. Image interpretation methods frequently consist of independent region growing segmentation followed by post-processing object labeling steps. This chapter considers another approach to image interpretation that integrates the segmentation and the object recognition steps into a single image interpretation process, incorporates contextual knowledge, and uses a genetic algorithm technique to produce an optimal image interpretation while utilizing a hypothesize-and-verify control strategy.

17 - Genetic Image Interpretation

View Online

Abstract

Details

Metrics