Home

Example how algorithm works

I use nice image as an example, because I am currently trying to optimize the algorithm for worse images.

Input:

Inputs are TEM images of gold nanoparticles and nanorods and .json file, which describes them.

Filtering:

I am using median filtering with kernel size dependent on image scale.

Binarizing:

I am using otsu thresholding method and combination of erosion and dilation for removing small holes and small objects (clearing image). I also use Canny edge detection and subtraction of these edges from binary image for better NPs separation.

Watershed and Hough transform

First the program performs watershed algorithm with seeds from distance map. HT is performed on regions where NPs are overlapping.