Shape Matching and Object Recognition Using Shape Contexts
IEEE Transactions on Pattern Analysis and Machine Intelligence
Hi-index | 0.00 |
This paper presents a new point matching algorithm which uses color pixel information in order to accomplish the registration task. Parameter estimation is carried out with a similarity transform estimated for sets of two pairs of points. The problem investigated in this paper is finding an optimal threshold that can reject possible mismatches between pairs of points, in order to simultaneously increase the estimation accuracy and reduce the processing time. The solution is analyzed and performances are evaluated.