Splitting an input image into connected sets of pixels is the purpose of image segmentation. The resulting sets, called regions, are defined based on visual properties extracted by local features. To reduce the gap between the computed segmentation and the one expected by the user, these properties tend to embed the perceived complexity of the regions and sometimes their spatial relationship as well. Therefore, we developed different segmentation approaches, sweeping from classical color texture to recent color fractal features, in order to express this visual complexity and show how it can be used to express homogeneity, distances, and similarity measures. We present several segmentation algorithms, like JSEG and color structure code (CSC), and provide examples for different parameter settings of features and algorithms. The now classical segmentation approaches, like pyramidal segmentation and watershed, are also presented and discussed, as well as the graph-based approaches. For the active contour approach, a diffusion model for color images is proposed. Before drawing the conclusions, we talk about segmentation performance evaluation, including the concepts of closed-loop segmentation, supervised segmentation and quality metrics, i.e., the criteria for assessing the quality of an image segmentation approach. An extensive list of references that covers most of the relevant related literature is provided.
CITATION STYLE
Ivanovici, M., Richard, N., & Paulus, D. (2013). Color image segmentation. In Advanced Color Image Processing and Analysis (Vol. 9781441961907, pp. 219–277). Springer New York. https://doi.org/10.1007/978-1-4419-6190-7_8
Mendeley helps you to discover research relevant for your work.