Active Shapes Models for Hand Posture Detection

Abstract

Image Segmentation is one of the prior steps to most of the image analysis tasks, especially if it relates to medical imaging. Among many other approaches, using statistical shapes models for image segmentation is a common practice in computer vision society. In this course work, we get introduced to one of these statistical shape models called Active Shape Model (ASM). ASM was implemented from scratch for hand shapes using the provided in-house data of hands having 40 examples with 56 landmarks for each. The main objectives of this project are as follows:

1. To understand the concept of Active Shape Model (ASM) and related theories to implement ASM from scratch using the provided data.
2. To understand and implement Procrustes Analysis for aligning, transforming (scaling and rotating) the different shapes to the reference shape.
3. To apply Principle Component Analysis (PCA) for reducing the dimensions of the feature space.