Color profile function for morphological image segmentation
Marcos Cordeiro d’Ornellas, Jose Antonio T. da Costa, Adriane Pedroso Dias · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 2005
In this paper, a new segmentation method is proposed. This method is based on the residuals of morphological opening and closing transforms with a geodesic metric. It may be considered analogous to a region growing technique. However, in contrast to using statistical local properties like in region growing approaches, the proposed method uses a pixel similarity rule based on the morphological characteristic of the connected components in the image. The proposed method is particularly well suited for the segmentation of complex image scenes such as aerial or fine-resolution images where very thin, enveloped and/or nested regions may have to be retained, and where the gradient calculation has a major drawback.