Bi-level multi-objective image segmentation using texture-based color features

This paper presents a bi-level multi-objective evolutionary image segmentation approach based on texture-based color features. There are many approaches for image segmentation. The majority of these approaches utilize either texture or color features of images. However, in most cases, only the texture or the color features are not sufficient for superior segmentation. Our proposed approach addresses this issue and employs a bi-level segmentation approach integrating both types of features. This approach uses color histogram-based texture feature in the first level of segmentation. In the second level, a multi-objective evolutionary algorithm is applied to the generated segmented images to produce the final set of non-dominated segmented results by optimizing three objectives simultaneously. The proposed approach is able to partition test images in a number of segments consistent with human visual perception. In term of quantitative evaluation, our approach also provides better results than existing approaches as shown in experimental results.