ASM

If you want to find or detect the object which has particular structures in some interesting images, active shape model is a good approach that you can try. It was proposed by Tim Cootes in 1995.

ASM is a “Model-Based” method, that is, it makes use of the prior statistical model of what is interesting or expected shape in the image, and then tries to find the best matching of the model to the target in the new images iteratively.

Statistical model is trained by some typical sample images and a set of landmarks which can describe the shape of the target and find on training images reliably. During the process of building the statistical model, in order to reduce the dimensionality of the shape data and simplify the problems, you should apply the Principle Component Analysis (PCA) to your aligned shape data. PCA computes the main axes of your shape data (2n-D space) and allowing one to approximate any of the original data using model with less than 2n parameters.

Finally, we could give a rough starting approximation shape and match to the target in new image using statistical model iteratively. More details about building the statistical model, PCA and matching processes you can find in the following references.

Active Shape Model is widely applicable, because the same algorithm can be applied to many different problems, just only by training with different training images.

Related Key technique words:
* Shape Alignment
* Principle Component Analysis
* Model Matching