On Using Physical Analogies for Feature and Shape Extraction in Computer Vision

Abstract

There is a rich literature of approaches to image feature extraction in computer vision. Many sophisticated approaches exist for low-and for high-level feature extraction but can be complex to implement with parameter choice guided by experimentation, but with performance analysis and optimization impeded by speed of computation. We have developed new feature extraction techniques on notional use of physical paradigms, with parametrization aimed to be more familiar to a scientifically trained user, aiming to make best use of computational resource. This paper is the first unified description of these new approaches, outlining the basis and results that can be achieved. We describe how gravitational force can be used for low-level analysis, while analogies of water flow and heat can be deployed to achieve high-level smooth shape detection, by determining features and shapes in a selection of images, comparing results with those by stock approaches from the literature. We also aim to show that the implementation is consistent with the original motivations for these techniques and so contend that the exploration of physical paradigms offers a promising new avenue for new approaches to feature extraction in computer vision. FEATURE EXTRACTION AND IMAGE PROCESSINGThis paper presents the first unified description of some new approaches to feature extraction in image processing and computer vision, derived by using physical analogies. The approaches aim to achieve feature extraction with insight for parameter selection, fast implementation by virtue of simplicity or by use of established techniques to improve computational speed, and with performance that is at minimum comparable with that achieved by state-or-art techniques. For this, we first explore what we understand to be meant by the term feature extraction. Feature and shape extractionThere is now a rich literature of techniques that can detect lowlevel features, such as edges and corners, and high-level shapes [1]. Low-level operators are generally those which operate on an image as a whole; high-level operators are those which process images so as to determine shapes that lie therein. Both processes are used within computer vision, to render explicit information that is implicit within the original image-as such providing image understanding. The state-of-art operators for low-level feature extraction include anisotropic diffusion for image smoothing, to preserve features and to reduce the effects of noise; the scale invariant feature transform (SIFT) aims to find corner features that persist over image scales, by a sophisticated operation based on the difference of Gaussian operator. By way of example, we show the result of anisotropic diffusion [2] applied to an image of an eye, in comparison with the result by Gaussian filtering, a standard operator. The anisotropic diffusion process in Fig. 1c achieves a more pleasing result than that of Fig. 1b, preserving features better while smoothing noise to greater effect. The parameter choice for the Gaussian oper...

Pressure Based Segmentation in Volumetric Images

“…We present a new 3D approach based on using a physical analogy [5] to separate intersecting regions. The approach is anisotropic and does not require a model or previous knowledge about the regions of interest.…”

Moving-edge detection via heat flow analogy

“…Results also show that this technique can detect moving-edges in image sequences without using the background image. Note that the preliminary version and the summary of this work were presented in (Direkoglu and Nixon, 2006;Nixon et al, 2009), respectively. In this paper, we extend the basis and evaluation of the new technique.…”

A novel physics-inspired method for image region segmentation by imitating the carrier immigration in semiconductor materials

“…It is still an on-going and open research topic to segment image regions for various practical purposes [3][4][5][6][7]. In recent years, nature-inspired methods have attracted more and more research attention, in which the mechanisms in nature are imitated and adjusted in novel algorithms for image segmentation, and promising results have been obtained in such preliminary works [8,9]. In this paper, inspired by the physical P-N junction, the self-balancing mechanism of carrier diffusing and drifting is adopted in a novel segmentation framework, in which the region structure of image is formed by dynamic carrier diffusing and drifting.…”