SKELETON PROCESSING FOR SHAPE ANALYSIS AND IMAGE GENERATION
Abstract
Thinning operations and distance transforms have long been used in extracting skeleton information of shapes. In this report, we propose some new algorithms to implement them and then study how they can be used to generate a structured descriptions of shapes. The feasibility of applying the structured shape descriptors for image generation is also demonstrated. The first algorithm proposed is an isotropic thinning algorithm for 2D binary images. The same idea is extended into 3D domain for the design of a new 3D thinning algorithm. A 3D isotropic thinning algorithm is also discussed. Then we propose a 3D Euclidean distance transform, its inverse transform, and the distance transform for 3D gray leveled images. Combining the above mentioned methods, a general algorithm for extracting reconstructable connected skeletons for both 2D and 3D images is described. The reconstructable connected skeletons form the basis of a structural shape descriptions for 2D and 3D objects. By spatial transformations on this skeleton-based models, a fast algorithm to generate the 2D projections of 3D objects, and a stochastic skeleton modeling scheme are then described.
Degree
Ph.D.
Subject Area
Electrical engineering
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.