Principle component analysis of images for the automatic location of control points

The method for establishing a relationship between at least two unlabelled images, comprising the steps of:

• A) conducting a training stage which includes the steps of:
• analyzing a plurality of representative labelled images to generate a model which identifies specific locations on one or more objects based on pixel values within each image;
• B) for each set of unlabeled images to be compared, conducting a model application stage which includes the steps of:
• matching each of said unlabelled images to the model to identify locations on objects within each unlabelled image;
• and
• determining correspondence between the unlabeled images based on the identified locations.

A method for establishing a relationship between at least two unlabelled images, comprising the steps of:

• A) conducting a training stage which includes the steps of:
• analyzing a first set of labelled images to generate a first model which identifies specific locations on one or more objects based on pixel values within each image of said first set;
• analyzing a second set of labeled images to generate a second model which identifies specific locations on one or more objects based on pixel values within each image of said second set;
• establishing a mapping between identified locations in said first model and identified location in said second model;
• B) for each set of images to be compared, conducting a model application stage which includes the steps of:
• matching a first unlabelled image to the first model to identify locations on objects in said first image;
• matching a second unlabelled image to the second model to identify locations on objects in said second image;
• and
• determining correspondence between the first and second unlabelled images based on the identified locations in each image and said mapping.

A method for indexing data into a multi-faceted model using a single-faceted global manifold model to which the data relates, comprising the steps of:

• A) conducting a training stage which includes the steps of:
• generating a single-faceted global manifold model from a plurality of examples of the data, wherein said global manifold model encompasses all examples of the data;
• generating a multi-faceted manifold model from divisions of the examples of data, wherein each facet encompasses a local set of examples of the data;
• B) for each new set of data, conducting a model indexing stage which includes the steps of:
• applying said global manifold model to the new set of data;
• deriving a coordinate vector from said application of the global manifold model to said set of data;
• and
• indexing said data into said multi-faceted model from said global manifold model in accordance with said coordinate vector.

A model for aligning observed data to a manifold model, comprising the steps of:

• i) selecting possible locations for the alignment of the data within an unaligned set of data;
• ii) for each possible location, determining a lower bound for the distance between the unaligned data set aligned at that location and an expected appearance of aligned data, in accordance with an average appearance defined by said model;
• iii) removing the possible locations whose lower bound exceeds a threshold value;
• iv) for each possible location, determining the coordinate value for a dimension of the model;
• v) for each possible location, determining a new lower bound by combining previously determined coordinate values with the distance between the data set aligned at that location and the appearance of the data set under said alignment in accordance with the previously determined coordinate values;
• and
• vi) repeating steps iii), iv) and v) for all of the dimensions of the model.