UNSUPERVISED FEATURE REDUCTION IN IMAGE SEGMENTATION BY LOCAL TRANSFORMS

We propose to reduce the dimensionality of feature vectors by using th e principles of Karhunen-Loeve transform (KL) applied to the feature i mages locally and globally. The reduction is achieved by choosing the resulting basis vectors which are closest to the classical KL basis ve ctors. An efficient implementation technique using pyramids is propose d. Experimental results are presented.