Probabilistic 2D Shape Retrieval and Applications via the Method of Hurwitz-Radon Matrices  

作　　者：Dariusz Jacek Jakobczak 

机构地区：[1]Department of Electronics and Computer Science, Technical University of Koszalin, Koszalin 75453, Poland

出　　处：《Journal of Control Science and Engineering》2014年第1期1-6,共6页控制科学与工程（英文版）

摘　　要：Artificial intelligence and computer vision need methods for 2D （two-dimensional） shape retrieval having discrete set of boundary points. A novel method of MHR （Hurwitz-Radon Matrices） is used in shape modeling. Proposed method is based on the family of MHR which possess columns composed of orthogonal vectors. 2D curve is retrieved via different functions as probability distribution functions： sine, cosine, tangent, logarithm, exponent, arcsin, arccos, arctan and power function. Created from the family of N-1 MHR and completed with the identical matrix, system of matrices is orthogonal only for dimensions N = 2, 4 or 8. Orthogonality of columns and rows is very significant for stability and high precision of calculations. MHR method is interpolating the function point by point without using any formula of function. Main features of MHR method are： accuracy of curve reconstruction depending on number of nodes and method of choosing nodes, interpolation of L points of the curve is connected with the computational cost of rank O（L）, MHR interpolation is not a linear interpolation.

关 键 词：Shape retrieval MHR coefficient of MHR method probabilistic modeling. 

分 类 号：O23[理学—运筹学与控制论]