An interactive algorithm for image denoising and segmentation
Marcos Carneiro de Andrade, Baba C. Vemuri · 2002
The paper presents an interactive algorithm for image denoising and segmentation. A global competition criterion is used to impose an order of processing on all image pixels. The smoothing step employs an evolution equation controlled by the local curvature to denoise the image while preserving the features. The interactive segmentation step requires the user to select one definitive seed per region. Region growing is initiated around provisory seeds, which are automatically detected, labeled and eventually merged by the algorithm. A simple merging mechanism is used to handle the topological transformations required to remove the image over-segmentation. It is shown that accurate and fast segmentation results can be achieved for gray and color images using this simple method. Extension to 3D images is straightforward and easily handled.