Comparison of evolutionary and swarm based computational techniques for multilevel color image thresholding

• Published in: Applied soft computing 2014-10, Vol.23, p.128-143
• Main Authors: Kurban, Tuba, Civicioglu, Pinar, Kurban, Rifat, Besdok, Erkan
• Format: Article
• Language: English
• Subjects: Color image thresholding; Evolutionary optimization algorithms; Swarm-based optimization algorithms
• ISSN: 1568-4946; 1872-9681
• DOI: 10.1016/j.asoc.2014.05.037

[Display omitted] •Multilevel image thresholding by using evolutionary algorithms has been investigated.•Algorithms used in the study are Evolution Strategy (ES), Genetic Algorithm (GA), Differential Evolution (DE), Adaptive Differential Evolution (JADE), Particle Swarm Optimization (PSO), Artificial Bee Colony (ABC), Cuckoo Search (CS) and Differential Search Algorithm (DS).•The Artificial Bee Colony Algorithm and Differential Search Algorithm are the most robust algorithms.•Differential Search Algorithm is the most effective algorithm in terms of CPU running times. This paper introduces the comparison of evolutionary and swarm-based optimization algorithms for multilevel color image thresholding problem which is a process used for segmentation of an image into different regions. Thresholding has various applications such as video image compression, geovideo and document processing, particle counting, and object recognition. Evolutionary and swarm-based computation techniques are widely used to reduce the computational complexity of the multilevel thresholding problem. In this study, well-known evolutionary algorithms such as Evolution Strategy, Genetic Algorithm, Differential Evolution, Adaptive Differential Evolution and swarm-based algorithms such as Particle Swarm Optimization, Artificial Bee Colony, Cuckoo Search and Differential Search Algorithm have been used for solving multilevel thresholding problem. Kapur's entropy is used as the fitness function to be maximized. Experiments are conducted on 20 different test images to compare the algorithms in terms of quality, running CPU times and compression ratios. According to the statistical analysis of objective values, swarm based algorithms are more accurate and robust than evolutionary algorithms in general. However, experimental results exposed that evolutionary algorithms are faster than swarm based algorithms in terms of CPU running times.