REFLECTANCE BASED OBJECT RECOGNITION

Neighboring points on a smoothly curved surface have similar surface n ormals and illumination conditions. Therefore, their brightness values can be used to compute the ratio of their reflectance coefficients. B ased on this observation, we develop an algorithm that estimates a ref lectance ratio for each region in an image with respect to its backgro und. The algorithm is efficient as it computes ratios for all image re gions in just two raster scans. The region reflectance ratio represent s a physical property that is invariant to illumination and imaging pa rameters. Several experiments are conducted to demonstrate the accurac y and robustness of ratio invariant. The ratio invariant is used to re cognize objects from a single brightness image of a scene. Object mode ls are automatically acquired and represented using a hash table. Reco gnition and pose estimation algorithms are presented that use ratio es timates of scene regions as well as their geometric properties to inde x the hash table. The result is a hypothesis for the existence of an o bject in the image. This hypothesis is verified using the ratios and l ocations of other regions in the scene. This approach to recognition i s effective for objects with printed characters and pictures. Recognit ion experiments are conducted on images with illumination variations, occlusions, and shadows. The paper is concluded with a discussion on t he simultaneous use of reflectance and geometry for visual perception.