A Hybrid Technique for Color Image Segmentation: Application to the Fire Forest Images

Salim Ben Chaabane, Moez Bouchouicha, Farhat Fnaiech

Abstract: This works deals with hybrid method for color image segmentation. The role of introducing the statistical features in clustering technique and that of data level and feature level fusion applied to color image segmentation are studied in this paper to obtain an optimal segmented image and to detect the fire in Forest images. The proposed segmentation approach is conceptually different and based on a new strategy. In fact, instead of considering an existing segmentation procedure, our technique rather explores the benefit of combining several approaches. However, the segmentation procedureis performed in two steps. In the first step, the segmentation of an image is obtained by integrating the statistical features and fuzzy clustering technique. For this propose, a modified Fuzzy c-means clustering is used to represent the information as fuzzy sets and to segment each image component into homogeneous regions. In the second step, on the obtained segmentation maps with specific primitive color, a combination rule and decision are employed to merge the segmentation results over different channels, in order to increase the quality of the information and to obtain an optimal segmented image. The classification accuracy of the proposed method is evaluated and a comparative study versus existing techniques is presented. The experimental results on synthetic and forest fire images demonstrate the value of integrating the statistical features in fuzzy clustering technique for image segmentation.

Keywords: Segmentation, Fire forest images, Fuzzy logic, Fuzzy C-Means, Fusion, first order statistical features