FAST OPTIMAL LINEAR APPROXIMATION OF THE IMAGES OF VARIABLY ILLUMINATED SOLID OBJECTS FOR RECOGNITION

摘要

An efficient computation of low-dimensional linear subspaces that optimallycontain the set ofimages that are generated by varying the illumination impinging on the surfaceof a three-dimensional object for many different relative positions of that object andthe viewing camera.The matrix elements of the spatial covariance matrix for an object arecalculated for an arbitrarypre-determined distribution of illumination conditions. The maximum complexityis reduced forthe model by approximating any pair of normal-vector and albedo from the setof all such pairsof albedo and normals with the centers of the clusters that are the result ofthe vectorquantization of this set. For an object, a viewpoint-independent covariancematrix whosecomplexity is large, but practical, is constructed and diagonalized off-line.A viewpoint-dependent covariance matrix is computed from the viewpoint-independentdiagonalizationresults and is diagonalized online in real time.