Hybrid Image Classification using ACO with Fuzzy Logic for Textured and Non-Textured Images

Background/Objectives: Classification is the most important tasks of decision making problems. It is used to group pixels into different groups in the image processing. It is frequently used to extract the land cover information, and to give a label to the area of the interest in the image. Methods/Statistical analysis: In this paper, a Hybrid classification approach, by combing the Ant Colony Optimization (ACO) and Fuzzy logic features is proposed. This approach is used to generate classification rules from the training set of the image. A measure of similarity for each pixel is calculated, which is almost same for the same class of the pixels with help of the proposed approach. Findings: It became a challenging task to classify textured and non-textured images in the presence of the coarse pixels values. The existing classification approaches such as statistical, knowledge-based and neural networks have many limitations in this context. In the training process the classification rules are generated. These rules are then given as input to the rule pruning process to further optimize the rules set. The generated classification rules are applied on the test set. It has been observed that the findings are providing better results even in the presence of mixed pixels. Application/Improvements: In this pixels of same image are being grouped into different groups. It can be extended further to apply this approach on the same category of the images for classification and Analysis.