Constructing a statistical shape model from two-dimensional or three-dimensional data

A computer implemented method of automatically parameterizing a set of three-dimensional shapes, each having a consistent topology, the method comprising automatically performing each of the following steps:

• (a) mapping each three-dimensional shape to a parameterized form of the consistent topology,
• (b) applying a set of landmarks to the parameterized form for a given level of resolution, each landmark being constrained to lie inside a spherical triangle formed by a triple of landmark points of a lower resolution level,
• (c) upon the completion of the application of the set of landmarks for the resolution level, projecting the set of landmarks onto the three-dimensional shapes,
• (d) building a statistical shape model using the set of landmarks of the resolution level,
• (e) using an objective function to provide an output which indicates a quality of the statistical shape model determined using the set of landmarks of the resolution level wherein the output of the objective function is a measure of the quantity of information required to code the set of shapes using the statistical shape model;
• (f) comparing the output with output determined using a different set of landmarks having the same level of resolution,
• (g) repeating steps (a) to (f) to select a set of landmarks based upon said output of the objective function,
• (h) repeating steps (a) to (g) at a higher level of resolution, and
• (i) outputting a selected set of landmarks as parameters corresponding to said three-dimensional shapes.

A computer implemented method of automatically parameterizing a set of three-dimensional shapes, each having a consistent topology, the method comprising automatically performing each of the following steps:

• (a) mapping each three-dimensional shape to a parameterized form of the consistent topology,
• (b) applying a set of landmarks to the parameterized form,
• (c) modifying the locations of the landmarks by applying a transformation,
• (d) projecting the set of landmarks onto the three-dimensional shapes,
• (e) building a statistical shape model using the set of landmarks,
• (f) using an objective function to provide an output which indicates a quality of the statistical shape model determined using the set of landmarks, wherein the output of the objective function is a measure of the quantity of information required to code the set of shapes using the statistical shape model;
• (g) comparing the output with output determined using landmarks having different locations,
• (h) repeating steps (c) to (g) to determine an output set of landmarks based upon said output of the objective function, and
• (i) outputting a selected set of landmarks as parameters corresponding to said three-dimensional shapes.

A computer implemented method of building a statistical appearance model by automatically establishing correspondence between a set of images, the method comprising:

• (a) determining a parameterization of each image, building a statistical model representing shape and appearance using the parameterization, using an objection function to automatically provide an output which indicates a quality of the statistical model, wherein the output of the objective function is a measure of the quantity of information required to code the set of images using the statistical model;
• (b) performing step (a) repeatedly for different parameterizations and comparing the quality of the resulting statistical models using output of the objective function to automatically select a parameterization based upon said output of the objective function;
• and
• (c) outputting the selected statistical model.