Browsing by Author "JOSHI, A"

Now showing 1 - 3 of 3

BACKPROPAGATION LEARNS MARR OPERATOR
(1993-11-01)
JOSHI, A
;
LEE, CH
;
資訊工程學系
;
Department of Computer Science
This paper describes a neural network model of the retinal responses to stimuli whose architecture is inspired by neurophysiological data. Suitable assumptions are identified which enable the development of a simple model for an individual X-type ganglion cell using backpropagation. This is then used to make a model of retinal processing. We present here our model of the individual ganglion cells and the underlying assumptions. We show that backpropagation leads to a model which is similar to the mathematical descriptions of retinal processing advanced by Marr. We present the results obtained when our model is used to simulate the effect of retinal processing on images. Empirical results about the speedups obtained when this model is implemented on parallel architectures are also reported.
ON CORRESPONDENCE, LINE TOKENS AND MISSING TOKENS
(1995-11-01)
LEE, CH
;
JOSHI, A
;
資訊工程學系
;
Department of Computer Science
In this article, the authors expand on some of their previous work on correspondence and propose a system that can correspond in the presence of both point and line tokens, or a mix of them. It can also work if some tokens present in one frame are not present in the other. The proposed algorithm is an amalgamation of reductionist and holist paradigms. We establish that properties which remain invariant for whole image sequences and sequences consisting of corresponding parts from each image frame can be used to establish correspondence. Extensive computer simulation results on synthesised data as well as real image sequences are presented to validate our claims.
ON THE PROBLEM OF CORRESPONDENCE IN RANGE DATA AND SOME INELASTIC USES FOR ELASTIC NETS
(1995-05-01)
JOSHI, A
;
LEE, CH
;
資訊工程學系
;
Department of Computer Science
In this work, the authors propose a novel method to obtain correspondence between range data across image frames using neural like mechanisms. The method is computationally efficient and tolerant of noise and missing points. Elastic nets, which evolved out of research into mechanisms to establish ordered neural projections between structures of similar geometry, are used to cast correspondence as an optimization problem. This formulation is then used to obtain approximations to the motion parameters under the assumption of rigidity (inelasticity). These parameters can be used to recover correspondence. Experimental results are presented to establish the veracity of the scheme and the method is compared to earlier attempts in this direction.