Low-Complexity Pruning for Accelerating Corner Detection

This is the implementation of Shi-Tomasi-Corner detection in low-complexity version, the method is from Low-Complexity Pruning for Accelerating Corner Detection,

the basic algorithm flow is

and you can refer the paper for more algorithm detail.

usage

mkdir build
cd build
cmake ..
cmake --build
cd ../prebuilt
./corner_detect xxx.jpg

and you will get the result xxx.jpg image under prebuilt/ folder

config.json

modify configuration as you need.

{
    "max_corner":500,
    "min_dist":20,
    "t1":0.01,
    "t2":0.01
}

Result

blox

There are  7791 corner candidate, 88.11 % corners be punned 
There are 2553 corners to be detected!
There are 49 corners after NMS

chess

There are  1692 corner candidate, 98.91 % corners be punned 
There are 1620 corners to be detected!
There are 108 corners after NMS

house

There are  9016 corner candidate, 86.24 % corners be punned 
There are 2287 corners to be detected!
There are 43 corners after NMS

lab

There are  49096 corner candidate, 81.27 % corners be punned 
There are 7457 corners to be detected!
There are 183 corners after NMS

Reference

• Low-Complexity Pruning for Accelerating Corner Detection

• An Analysis and Implementation of the Harris Corner Detector

• https://github.com/coder-mano/Shi-Tomasi-Corner-Detector