hi
I need matlab code for elastic bunch graph matching. I have 2 faces ,I should making their mesh and compare them.I think elastic bunch graph matching can help me with it.

introduction
Our goal is to recognize individuals from the individual images by clicking on the Gallery, which also lists only one image per person. Our task is to deal with changing the image of the differences in the expression of the face, head pose, position and size (to name only the most important). Our task, therefore, is a typical discrimination-in-presence-variance problem, where it is necessary to try to minimize dispersion and stress the discriminatory features. This is normally only possible by using information about the structure of the system changing the direction expected.The Classification is significantly different in nature and origin of their knowledge about the image
variations. Systems of artificial intelligence and computer vision often emphasize the specific design provided by structure, for example, explicit models of 3D objects or process image generation, while neural network model tends to emphasize the absorption structure on examples using statistical estimation methods. Both of these extremes in its own way and fall painfullyshort the ease with which natural systems to pick up information from just a few examples. Part of the success of natural systems should be linked to the particular characteristics and laws on how to convert the picture object under natural conditions. Our system is an important foundation of a structure that reflects the fact that the images are coherent objects generally to translate, scale, rotate, and warp in the image plane. Our main object view tagged Earl; the ribs which are labeled with information about the distance and the nodes labeled Wavelet responses
on the set of jets. Retained graphics model can be mapped to the new images to obtain images of charts, which can then be included in the Gallery, and become a model of graphics. Bursts, as we use them resistant to moderate changes in lighting and small displacements and deformations. Model graphics can be easily translated, scaled,
oriented or deformed during reconciliation, therefore, compensate for most of the variance in the images. Unfortunately, having only one image for each person in the galleries do not give sufficient information to process the rotation in the same way. Nevertheless, we present results on the recognition of different poses.