1. Reading original image using SciPy.
  2. Cropping the image to the lower half (to remove non-bike lane features), and clip the bottom 10% of the picture as well (to remove bike tire from the view).
  3. Filtering — In this case, a median filter is applied to preserve edges.
  4. Converting the image from RGB to HSV and filter image to preserve only darker (defect) pixels.
  5. Edge detection using the canny edge detection.
  6. A 3x3 gaussian filter is applied over the image and then perform erosion / dilatation operations on remaining white pixels to remove noise.