This repository contains several MATLAB functions and a demo script we used to detect needle shaped crystals on images for monitoring their growth. The results and description of the method can be found in [1]. Please reference this publication if you use this code for your research.
The method is based on MATLAB's Radon transform based line detector with several improvements in order to increase its perfromance and quality for the images with large amount of line segments. The main imrpovements are following:
- Splitting image into segments (using
bwlabel()) and processing the segments separately - Using Gaussian blur to remove a "baseline" for better peak detection in Rho/Tau space
- Possibility for fine tuning of the algorithm
The main function is getlinesforbw.m, it takes a binary image as a main argument
and returns a structure array with detected lines. You can also provide a structure
with parameters as a second argument, see demo.m for example and more details.
Notice, that:
- Changing some of the parameters (e.g.
theta_step) increases computational time. - Optimization of the paremeters using e.g. DoE approach is a good idea.
The scripts require Image Processing Toolbox.
- Image Analytical Approach for Needle-Shaped Crystal Counting and Length Estimation. Wu, Jian X.; Kucheryavskiy, Sergey V.; Jensen, Linda G.; Rades, Thomas; Müllertz, Anette; Rantanen, Jukka. Crystal Growth & Design, Bind 15, Nr. 10, 2015, s. 4876-4885.