Local curvature constrained Level Set segmentation using a Spectral Bi-Partitioning algorithm
Fahima Djabelkhir, Karim Mokrani
Abstract
Fahima Djabelkhir, Karim Mokrani
Abstract
Coupling the level set method and graph cut optimization method in a complementary fashion demonstrates a superior performance in two-class segmentation problems. The basic idea is to first define an energy function according to curve evolution using Level Set method and then construct a similarity graph with well selected edge weights based on the boundary curvature values, which is further optimized via our proposed spectral bi-partitioning algorithm. By the way, our model shares advantages of both level set methods and graph cut algorithms. Difficulties are found in computing due to the fact that curvature values are very critic and not easy to manipulate, start working with those values leads to instability in computing. To avoid this problem, we have chosen images with two-class segmentation problems. Accordingly, the proposed method uses local criteria and yet produces results that reflect global properties of the image.
A significance statement is not available in the OpenAlex record.
A contribution statement is not available in the OpenAlex record.
Method details are not available in the OpenAlex metadata.
Findings are not separately available in the OpenAlex metadata.
Limitations are not available in the OpenAlex metadata.
Application details are not available in the OpenAlex metadata.
Coupling the level set method and graph cut optimization method in a complementary fashion demonstrates a superior performance in two-class segmentation problems. The basic idea is to first define an energy function according to curve evolution using Level Set method and then construct a similarity graph with well selected edge weights based on the boundary curvature values, which is further optimized via our proposed spectral bi-partitioning algorithm. By the way, our model shares advantages of both level set methods and graph cut algorithms. Difficulties are found in computing due to the fact that curvature values are very critic and not easy to manipulate, start working with those values leads to instability in computing. To avoid this problem, we have chosen images with two-class segmentation problems. Accordingly, the proposed method uses local criteria and yet produces results that reflect global properties of the image.
Key concepts: Curvature, Image segmentation, Algorithm, Level set (data structures), Cut, Computer science, Graph, Segmentation