Method and apparatus for object normalization using object classification

A method for classifying an object, comprising:

• obtaining a set of classification results for a given object class across a plurality of sequential images for a plurality of viewpoints, each of said classification results comprised of a position of one of said objects in said sequential image, and at least one projected property of said object at said position;
• determining normalization parameters for each of said plurality of viewpoints by fitting a high order model to said classification results;
• computing normalized training data during a training mode;
• and
• classifying said object using a set of normalized features and said normalized training data.

An apparatus for normalizing objects across a plurality of image viewpoints, the apparatus comprising:

• a memory;
• and
• at least one processor, coupled to the memory, operative to:
• obtain a set of classification results for a given object class across a plurality of sequential images for each of said plurality of viewpoints, each of said classification results comprised of a position of one of said objects in said sequential image, and at least one projected property of said object at said position;
• and
• determine normalization parameters for each of said plurality of viewpoints by fitting a high order model to said classification results to model a change in said at least one projected property.