Most image processing architectures adapted to morphological operations use structuring elements of a limited size. Various algorithms have been developed for decomposing a large sized structuring element into dilations of small structuring components. However, these decompositions often come with certain restricted conditions. In this paper, we present an improved technique using genetic algorithms to decompose arbitrarily shaped binary structuring elements. The specific initial population, fitness functions, dynamic threshold adaptation, and the recursive size reduction strategy are our features to enhance the performance of decomposition. It can generate the solution in less computational costs, and is suited for parallel implementation.
All Science Journal Classification (ASJC) codes
- Signal Processing
- Computer Vision and Pattern Recognition
- Genetic algorithms
- Mathematical morphology
- Structuring element