Quantum Inspired Swarm Optimization for Multi-Level Image Segmentation Using BDSONN Architecture

Quantum Inspired Swarm Optimization for Multi-Level Image Segmentation Using BDSONN Architecture

Subhadip Chandra (Camellia Institute of Technology, India) and Siddhartha Bhattacharyya (RCC Institute of Information Technology, India)
Copyright: © 2015 |Pages: 41
DOI: 10.4018/978-1-4666-8291-7.ch009
OnDemand PDF Download:
List Price: $37.50


This chapter is intended to propose a quantum inspired self-supervised image segmentation method by quantum-inspired particle swarm optimization algorithm and quantum-inspired ant colony optimization algorithm, based on optimized MUSIG (OptiMUSIG) activation function with a bidirectional self-organizing neural network architecture to segment multi-level grayscale images. The proposed quantum-inspired swarm optimization-based methods are applied on three standard grayscale images. The performances of the proposed methods are demonstrated in comparison with their conventional counterparts. Experimental results are reported in terms of fitness value, computational time, and class boundaries for both methods. It has been noticed that the quantum-inspired meta-heuristic method is superior in terms of computational time in comparison to its conventional counterpart.
Chapter Preview


Image segmentation is one of the most important challenges encountered in the field of image processing (Gonzalez & Woods, 2002), which is crucial for image understanding and analysis to interpret its contents. The objective of image segmentation is to extract meaningful non-overlapping homogeneous regions from an image. The process of image segmentation is executed based on the principle that each of the pixels in a region is similar to other with respect to some characteristics such as intensity, texture or color. Segmentation can be carried out by several classical techniques, viz. histogram based, edge based, region based split/ merge techniques (Ho & Lee, 2001). Histogram based approaches are those in which pixels are classified using the histogram of the images according to their color intensity. Pixels representing marked intensity shifts are extracted and then linked into contours that represent object boundaries are offered in edge based approaches. These approaches offer low computational cost but on the other hand pose serious difficulties in setting the appropriate thresholds and producing continuous one-pixel-wide contours (Sahoo, 1988; Helterbrand, 1996). Region based approaches aim to detect regions satisfying a certain homogeneity criterion. This class includes region growing (Adams, 1994; Chang, 1994; Hojjatoleslami, 1998) and pyramidal methods (Rezaee, Van der Zwet, Lelieveldt, van der Geest, & Reiber, 2000) which are powerfull but may lead to an over segmentation. Split/merge approaches aim to overcome the problem of over segmentation by means of a two phase process. The first phase subdivides the original image into primitive homogeneous regions. The second one tries to get a better segmentation by merging neighboring regions which are judged similar enough (Chun, 1996; Bhandarkar, 1999).

The objects in an image usually have a strong correlation with the regions of the segmented image. The resulted segmented image is labeled in such a way that facilitates the description of the original image so that it can be interpreted by the system that handles the image. To determine which are the features that can lead to successful classification, a priori knowledge or/and presumption about the image are generally needed. Most of the image segmentation algorithm yield segmentation of different objects with respect to the image background.

Here both methods are capable to perform multilevel image segmentation of gray scale images. The soft computing approaches applied in this direction either resorts to a deterministic analysis of homogeneous intensity values of images or to an application of heuristic search and optimization techniques, though these techniques suffer from several degrees of random time complexity.

Swarm intelligence (Englebrecht, 2002), which takes inspiration from the social behavior of insects and other animals, is a relatively new computational approach to solve problems. In a PSO system, particles fly around in a multidimensional search space. During flight, each particle adjusts its position according to its own experience, and the experience of its neighboring particles, making use of the best position encountered by itself and its neighbors.

Key Terms in this Chapter

BDSONN: This architecture is a three-layer network structure assisted by bi-directional propagation of network states for self-supervised organization of input information.

Particle Swarm Optimization: It is a computational method that optimizes a problem by iteratively trying to improve a candidate solution with regard to a given measure of quality.

Image Segmentation: It is the process of partitioning a digital image into multiple segments on the basis of their similarities.

Optimization: Optimization is the process of finding the greatest or least value of a function for some constraint, which must be true regardless of the solution. In other words, optimization finds the most suitable value for a function within a given domain.

Swarm Intelligence: Swarm intelligence is the collective behavior of decentralized, self-organized systems, natural or artificial.

Ant Colony Optimization: Ant colony optimization (ACO) is a population-based metaheuristic that can be used to find approximate solutions to difficult optimization problems. In ACO, a set of software agents called artificial ants search for good solutions to a given optimization problem.

Quantum Computing: Quantum computing is the area of study focused on developing computer technology based on the principles of quantum theory, which explains the nature and behavior of energy and matter on the quantum level.

Complete Chapter List

Search this Book: