masm: Manchester's Active Shape Model library

The masm library contains classes to implement a 2D Active Shape Model (ASM).
The ASM is an algorithm to match statistical shape models to image - or rather, to find a set of model points on an image with positions constrained by a statistical shape model.
It comes with a set of tools to build ASMs and to run them over sets of images (see the tools directory).

Requirements
The Basic Algorithm
Example Usage
Sequences of ASMs

Requirements

The masm library depends on VXL, and in particular contrib/mul/msm - the statistical shape modelling library in Manchester's contributions to VXL.

The Basic Algorithm

Given a shape model instance defining the initial position of a set of model points on an image, the basic ASM algorithm iterates the following steps:

Search around each point with a local model to find the best local position of that point
Optimise the shape and pose parameters to best fit to the new point positions

The masm_model class contains the data required for an ASM. In particular:

A copy of a statistical shape model (msm_shape_model)
A set of msm_point_finder_model objects, one for each point.

To search an image with an ASM, you must create a masm_searcher object, which contains the current shape and pose parameters, as well as a set of msm_point_finder objects, which will do the search for each point.
There can be multiple masm_searcher objects, all using the same masm_model.

Example usage

To load in a model and use it to search an image: // Load in the ASM masm_model model; if (!vsl_quick_file_load(model,params.asm_path)) return 1; // Create a searcher (which will be initialised with the default pose) masm_searcher asm_searcher(model); // Load in an image vimt_image_2d_of image; vimt_load(image,image_path); if (image.image().size()==0) return 2; // Most models assume grey-scale images. If RGB then convert to grey. if (image.image().nplanes()==3) { vil_image_view grey_image; vil_convert_planes_to_grey(image.image(),grey_image); image.image()=grey_image; } // Build an image pyramid from this vimt_gaussian_pyramid_builder_2d pyr_builder; vimt_image_pyramid image_pyr; pyr_builder.build(image_pyr,image); // Perform the search asm_searcher.search(image_pyr); // Get the resulting points msm_points final_points = asm_searcher.points();

Sequences of ASMs

It is often useful to use a sequence of ASMs to get faster and more robust results.
For instance, the first ASM may work at a low resolution with only a few shape parameters - this can get to the approximate solution quickly. Subsequent ASMs will then work at increasing resolution with more shape modes. This is a generalisation of the original "Multi-Resolution ASM" idea.
This is expressed in the library using the masm_model_series class, which stores a sequence of masm_model objects, and the associated masm_series_searcher which stores a set of masm_searcher objects, which will be applied in sequence.
The latter is used in almost exactly the same way as the single ASM: // Load in the ASM sequence masm_model_series model; if (!vsl_quick_file_load(model,params.asm_path)) return 1; // Create a searcher (which will be initialised with the default pose) masm_series_searcher asm_searcher(model); // Perform the search asm_searcher.search(image_pyr); // Get the resulting points msm_points final_points = asm_searcher.points(); For working examples of using the classes, see the programs in the tools and tests subdirectories.