souravc83 / pbtspot

Probabilistic Boosting Tree for Spot Detection in Fluoroscence microscopy images

pbtspot

A python library which uses Probabilistic Boosting Tree for Spot detection in fluoroscence microscopyy images.

Background

Spot Detection in fluorosence microscopy images is complicated, due to low signal to noise ratios. Machine learning methods are known to perform better than traditional image processing techniques in these situations. This project uses a probabilistic boosting tree, which is a decision tree. Each node of the decision tree is a strong Adaboost classifier (a linear combination of a set of weak Haar-like features). Such classifiers are used widely in face recognition, following the celebrated paper of Viola and Jones. The main goal of this project was to improve upon the Viola-Jones classifier using a probabilistic method, which can then be used subsequently for Bayesian tracking methods, which have become popular in fluoroscence particle tracking in the recent past.

Installation

python setup.py build
sudo python setup.py install 
(On a windows system, simply use "python setup.py install")