Perceptual image analysis

Abstract

The problem considered in this paper is one of extracting perceptually relevant information from groups of objects based on their descriptions. Object descriptions are qualitatively represented by feature-value vectors containing probe function values computed in a manner similar to feature extraction in pattern classification theory. The work presented here is a generalisation of a solution to extracting perceptual information from images using near sets theory which provides a framework for measuring the perceptual nearness of objects. Further, near set theory is used to define a perception-based approach to image analysis that is inspired by traditional mathematical morphology and an application of this methodology is given by way of segmentation evaluation. The contribution of this article is the introduction of a new method of unsupervised segmentation evaluation that is base on human perception rather than on properties of ideal segmentations as is normally the case.