Harris-Corner-Detection using Sobel and Roberts Cross operator

Installation:

Requirements:

• OpenCV 4.0.0 [http://opencv.org/]
• g++ 7.4.0
• OS Linux

How to build:

• cd harrisCornerDetect
• cd Debug
• make clean
• make all
• It will build the binary execution file in the Debug directory. To run the binary file please follow the following instructions.

Execution:

In the Debug directory

• ./harrisCornerDetect image-file Gaussian-Std NMS-threshold NMS-window-size NMS-window-separation Grad-sum-window-size gradient-type
• Image-file (string): Complete path of the jpg image file (color or grayscale). Following input parameters are optional.
• Gaussian Std (float): Standard deviation of the Gaussian filter. The value must be greater than zero. Default: 1.0
• NMS threshold (float): Threshold applied to the maximum Harris response value. This value must be between 0 and 1. Default: 0.02
• NMS window size (int): Size of the square window in the response matrix where the local maxima are computed. The value must be greater than zero and must be smaller than the smaller dimension of the input image. Default: 3
• NMS window separation (int): The gap between two subsequent NMS windows. The value must be greater than zero and less than non-maximum suppression window size. Default: 1
• Grad Sum window size (int): Size of the square window of the gradient matrix to compute the summation of products. The value must be greater than 1 and odd number and must be smaller than the smaller dimension of the input image. Default: 3
• gradient type (int): 1 is for Sobel and 2 is for Roberts gradient. Default: Sobel.Output:
• The updated image (i.e., out.jpg) with corner locations will be saved in the debug directory.

Results:

Original image:

Using Sobel operator:

Using Roberts operator:

Performance Analysis: