	LOGIN
	>> Home
	>> Topics
	>> Students
	>> Partners
	>> Statistics

Information for topics

Topic Id:

ID topic:	586
Partner Email:	dtdim@iinf.bas.bg
Project Title:	Exact Computer Performance of Direct and Inverse Hough/Radon Transform
Abstract:	Hough Transform (HT), patented in USA in 1961, was expanded during the 80-ies of the 20-th century by a set of modifications serving as an appropriate tool in images processing. Some of the most famous are the so called HT modifications, such as (,)-HT of Duda and Hart (1972), detecting linear segments in images, as well as HT-modification of Ballard (1981) for arbitrary curves, and others. At the same time (rho,theta)HT is an isomorphic version of 2D Radon Transform (RT) (1917), known also as X-ray transform. The multidimensional expansions of RT are undoubtedly useful in mathematics, but RT popularity comes mostly thanks to its computer tomography application, where inverse RT is used. In fact, tomography is a very convenient method to obtain precise 3D models of real objects, or at least of objects, which efficiently let through “Х-rays”. The Institute of Information and Communication Technologies is working on a development of the so called “exact realization of (rho,theta)HT under computer resources constraints”. This project realizes structural approximation of the concept of “exact (rho,theta)-HT algorithm”. Attention is also paid to the direct (and obviously faster) performances. A similar principle might also be used in the realization of the inverse (rho,theta)HT. Thus the interest towards the development is motivated in three aspects: direct (rho,theta)-HT for the purposes of images classic processing; Inverse RT -as tomography application; and the two latter – for classic recognition of 2D/3D objects. Purpose of the diploma work offered: (1) To experiment an analytical solution of the exact performance of (rho,theta)-HT under computer resources constraints; (2) On a similar principle to develop the inverse (rho,theta)-HT. To accomplish a comparative experiment with the help of the method “filtered inverse projection” for realization of 2DRT-1; (3) To describe in details НТ in some well known sites, such as Wikipedia, WolframMathWorld and others. Not limiting the generality, it is assumed that purpose (1) would attract mainly the attention of post graduates in “Computer Graphics” specialties, while purpose (2) – those in “Biomedical informatics” specialties. Purpose (3) is optional, attached to (1) and/or (2), when specific interest is shown by the student. In addition to the theoretic part, the development must include computer experiments containing: - software modules for realization of one of the two purposes - (1) or (2); - a test system for demonstration of the modules developed. Initial information concerning the diploma work could be found in the paper: Dimov D. Using an Exact Performance of Hough Transform for Image Text Segmentation, In: Proceedings of Int. Conference on Image Processing (ICIP’2001), Oct. 7-10, 2001, Thessaloniki, Greece, Vol. I, pp.778-781. Recommended environment for the software experiments – Borland C/C++ Builder.
Advisor:	Assoc. Prof. Dr Dimo T. Dimov
Link:
Degree:	Master
Keywords:
Artificial intelligence & Neural networks
Image processing