Combining Contourlet Transform and Watershed for Robust Multi-Scale Image Segmentation

Abstract

This paper proposes a new technique’s segmentation that combines multiresolution contourlet transform with the watershed transform. The contourlet transform is one kind of new multi-scale transform that is based on discrete curvelet transform, , whose structural elements include the parameters of location and dimension, and orientation parameter more, which let contourlet transform has good orientation characteristic. Therefore, contourlet transform is superior to curvelet in the expression of image edge, and denoising such as geometry characteristic of discrete curve and beeline, which has already obtained very good research results in image denoising. The flooding watershed transform is then applied, and the segmented image is projected up to higher resolutions using contourlet transform. Typically, if a low resolution is chosen for the initial segmentation so large relevant objects will be captured (noise); therefore, a higher initial resolution will lead to smaller and more detailed segmented objects. This paper puts forward an improved method based on contourlet and watershed transform because certain regions of the image have the ringing and radial stripe after contourlet transform.