Recognizing Objects in Cluttered Images using Subgraph Isomorphism ∗

This paper reports on a new approach to object recognition based on exploiting both geometric and spectral information in an image. An algorithm is described for finding objects in an image based on inexact graph matching against a template that incorporates information about the geometry and color of segmented regions of the image. The image is encoded as an attributed graph in which vertices represent regions, and are annotated with the position, shape and color of the image. Finding the template in a new image takes place in three steps: local, neighborhood and global matching. In the last step a maximal set of mutually compatible template / candidate image region assignments is sought. Currently the system has been evaluated using 3-band color images, though experiments carried out with 4-band images suggest improved performance with such images.

[1]  Yoshiaki Shirai,et al.  Three-Dimensional Computer Vision , 1987, Symbolic Computation.

[2]  Bir Bhanu,et al.  Segmentation of natural scenes , 1987, Pattern Recognit..

[3]  Azriel Rosenfeld,et al.  Computer Vision , 1988, Adv. Comput..

[4]  James A. McHugh,et al.  Algorithmic Graph Theory , 1986 .

[5]  Fuyau Lin,et al.  A parallel computation network for the maximum clique problem , 1993, 1993 IEEE International Symposium on Circuits and Systems.

[6]  Marcello Pelillo,et al.  Relaxation labeling networks that solve the maximum clique problem , 1995 .

[7]  Steven Gold,et al.  A Graduated Assignment Algorithm for Graph Matching , 1996, IEEE Trans. Pattern Anal. Mach. Intell..

[8]  Glenn Healey,et al.  Computing illumination-invariant descriptors of spatially filtered color image regions , 1997, IEEE Trans. Image Process..

[9]  Glenn Healey,et al.  Exploiting an atmospheric model for automated invariant material identification in hyperspectral imagery , 1998, Defense, Security, and Sensing.