Perception‐based image classification

Abstract

Pattern classification methodologies are present in many systems that we depend on daily. In these systems, classes are created based on human perception of the objects being classified. Thus, it is important to have systems that accurately model human perception. Near set theory provides a framework for measuring the similarity of objects based on features that describe them in much the same way that humans perceive objects. In this paper, we show that the near set approach can be used to classify images. Further, the results presented here suggest that the near set approach can be used in any image classification system. The contribution of this article is a perception based classification of images using near sets.

Plastic Surgery Facial Image Correspondence: A Near Set Approach

“…Specifically, the nearness measure was introduced by Henry and Peters in [18]. The nearness measure presents a systematic approach to determining the degree of similarity between a pair of disjoint sets, an idea that can be visualized by asking "which pair of sets in Fig.2 are more similar?"…”

“…This process is then repeated for the set Y because the two distances will not necessarily be the same. Keeping this in mind, the measure tHD [18] is defined as…”

Anisotropic Wavelet-Based Image Nearness Measure

“…Each image feature is represented by what is known as a probe function, a partial model of perception viewed as a mapping φ : X → ℜ inspired by a psychophysics interpretation of the relation between a set of stimuli X and sensation [1]. The notion of an image probe function was first introduced in 1993 by M. Pavel as part of a study of image registration [29] and later refined in [16]. The basic idea is to "probe" images as part of a feature-extraction process.…”

“…Kerre [57] on morphological image interpolation in magnifying colour images with sharp edges. A direct benefit of this research is an effective means of grouping together (classifying) images that correspond to each other relative to minuscule similarities in the contour, position, and spatial orientation of bounded regions in the images, especially in videos containing image sequences showing varied object movements (see, e.g., [15,16,40,42,44]). The contribution of this article is the introduction of an anisotropic waveletbased measure of image resemblance using a near set approach.…”

“…The proposed approach to measuring image similarity stems from recent work on anisotropic wavelets [46][47][48], image correspondence [13,16,18,37,39] and near sets [16, 26, 34-37, 39, 44]. Similarities between images can be measured by taking into account image features such as contour, spatial orientation and position of line segments along bounded regions contained in sample images.…”