Trabecular Bone Image Segmentation Using Iterative Watershed and Multi Resolution Analysis

Usually, bone fragility risk is related to deteriorations of osseous architecture. However, medical imaging is one of the means to appreciate in vivo bone screen, such as microscopic or micro-tomography images, which is important in the follow up of the osteoporosis. In this paper, a new image segmentation technique of trabecular bone images is introduced. It combines both hierarchical watershed segmentation, wavelet and image mosaic transform. The wavelet transform is applied to the intensity image, to de-noise the image, enhance edges in multiple resolutions, creating detail and approximation coefficients. Gradient magnitudes of the approximation image at the coarsest resolution are computed. The hierarchical watershed and the image mosaic transform are then applied to the approximation image at a given resolution. The segmented image is projected up to higher resolutions using the inverse wavelet transform. This technique provides robust segmentation results for images; reduces the watershed algorithm over-segmentation and results in closed homogeneous regions.