Region-extraction-from-an-RGB-image

Summary

A sample program for extracting a specific region from an RGB image using openCV and python. Bounding-box information, actual contour information and other physical parameters such as centroid, area… etc are also calculated.

Intial RGB image is uploaded. (By default instead of RGB, its BGR)

RGB image is converted to grayscale image

The grayscale image is undergoes blurring using “GaussianBlur()” functionality.

The blurred image is used for detecting edges. “Canny()” function is used for this purpose.

To remove unwanted edges and close the gaps, morphological transformations using OpenCV are applied. Here the canny image undergoes two transformations.

At first, the image undergoes a closing operation and secondly, the same canny image undergoes dilation and erosion. By using 'and' operator into these results provide effective removal of unwanted noises.

Floodfill functionality is used to fill up the void as shown below:

The output undergoes thresholding and using "findContours" functionality, the contours can be found out.

Bounding rectangle is detected for each contour using OpenCV “boundingRect()”.

The actual countour of the region can also be saved using "imutils.grab_contours()" functionality.

Final ROI is saved separately along with other physical parameters.

By saving ROI in this manner can be used to create a database for generating synthetic image dataset