The non-parametric Parzen's window in stereo vision matching

Abstract

This paper presents an approach to the local stereovision matching problem using edge segments as features with four attributes. From these attributes we compute a matching probability between pairs of features of the stereo images. A correspondence is said true when such a probability is maximum. We introduce a nonparametric strategy based on Parzen's window (1962) to estimate a probability density function (PDF) which is used to obtain the matching probability. This is the main finding of the paper. A comparative analysis of other recent matching methods is included to show that this finding can be justified theoretically. A generalization of the proposed method is made in order to give guidelines about its use with the similarity constraint and also in different environments where other features and attributes are more suitable.

A Probabilistic Neural Network for Attribute Selection in Stereovision Matching

“…This is the case when the features are regions and the attributes are the set of seven Hu invariant moments by example. In earlier work [20] we have carried out comparative analyses between different local learning matching strategies, and the non-parametric Parzen's window estimation approach is compared with the PNN proposed in this paper [21]. We have also included both PNN and Parzen's window in global matching strategies [21] as a mapping of the similarity constraint competing with the smoothness and uniqueness global matching constraints.…”

An explicit and compact coding of geometric and structural image information applied to stereo processing

“…The aim of this paper was to define and investigate a local multi-modal similarity function. A combination of a local matching with more global ordering or smoothness constraints (as for example in Koch, 1994;Pajares and de la Cruz, 2002) or the utilisation of multiple time frames (see, e.g., Schmid and Zisserman, 1997;Kr€ uger et al, 2002b) can be used to further improve performance.…”

Fuzzy Cognitive Maps for stereovision matching

“…We have studied the performance of some local matching strategies: Fuzzy Clustering [12], Support Vector Machines [13], Hebbian learning [16] or a nonparametric probabilistic approach [17]. Nevertheless, we have also verified that the best performance in stereovision matching is achieved under global matching strategies [6][7][8].…”

“…The following papers use a global relaxation technique based on probabilistic/merit [8,[17][18][19][20][21][22][23][24][25] and optimization through a Hopfield neural network [4,6,28,29].…”