CV-workshop5

Workshop 5

MISCADA – Computer Vision

Prof. Paolo Remagnino

Colour detection and segmentation

• Go to
  • https://realpython.com/python-opencv-color-spaces/
• Read the article
• Use the proposed method to segment an image
  • Some images have been posted in the Blackboard folder for workshop5
• Select an area of a given colour, for instance a rectangular area
• The colour of the pixels in that region can be employed to create the mean and standard deviation for the region colour
• Use that information to write a simple algorithm that detects pixels of that colour
• Plot the results

$P={(s_x,sy),…(e_x,ey)}$

Example

• You compute mean and standard deviation of P
• Try to do this with a region in the sky
• You then use the mean and deviation to select in the image those pixels that fit that Gaussian
• You could even create a heatmap where the expected colour is

Colour segmentation

• With any colourful image
  • Segment in the (r, g) space, where r and g and the normalized red and green components
  • Segment in the (u, v) space, after having converted pixels from RGB to YUV

Hough Transform

• Go to
  • https://docs.opencv.org/3.4/d9/db0/tutorial_hough_lines.html
• Learn about the Hough transform and how to use it in OpenCV
• Experiment with the chess and go board images I have provided in the same folder
• Try now tuning all the tunable parameters
• Plot all the results