Count Bro

Program written in C using no additional libraries, providing a toolset used to count objects where no overlapping occurs (e.g. cells, coins) and specific constraints are set. (mainly based on colour)

The program implements several image-processing tools which can be used independently. The intended workflow for counting objects is:

1. Threshold the input image, using a well-chosen thresholding interval
2. Apply the necessary morphological operations (dilation, closing, opening, erosion)
3. Use the labelling algorithm and count the objects

The project is made with no additional libraries, being created with a learning purpose in mind.

Features:

1. BMP file interface, able of I/O and general access
2. Implementations of data structures such as vector, set
3. Interfaces for various image types such as binary or grayscale
4. Implementation of morphological operators, used for detail enhancing or impurity removal
5. Basic image processing specific kernel functionality
6. Colour based two-pass labelling algorithm

Installation

Simply clone the project through your preferred IDE or by using the bash command

git clone git@github.com:thesstefan/countbro.git

Usage

After closing the repository, write your customized main.cpp (see some examples in extra/) and modify the Makefile to suit your needs (specify input/output image file paths). Next, compile the project and count the objects in the image by executing (assuming a shell environment)

make -jX && make run

where X is the desired number of jobs to be ran in parallel.

Examples:

1. Counting (1033) and detection of stars using extra/main_sky.c as source and data/sky_input.bmp as input.

2. Counting (10) and detection of some coins using extra/main_coins.c as source and data/coins_input.bmp as input.

3. Counting and detection (~184 with overlapping cells errors) of blood cells in a scan (the original purpose of the project), using extra/main_cells.c as source and extra/data_input.bmp as input.

Future Plans:

1. Implement colour-range thresholding for finer detection control
2. Transition everything to shader & GPU implementations for performance improvements

Icons made by Freepik from www.flaticon.com