3

So which do you think would be faster to detect related points in an image to a certain threshold ?

A. Scan a line at a time and define a rectangle surrounding the shape ?

B. Starting at a pixel find values in each direction within a tolerance and recurse each point found with the same function

C. Do something similar to above but try to find the edges by finding the last point before blank space to get a shape

D. Identify all line segments on the horiz very and diagonals and see which ones intersect ? Omg I asked this before. After discovering all the points that are within threshold and iterating through these alone?

E. Is there another goddamn method ??? Lol

Comments
  • 0
    @rantsauce eh ? Any thoughts ?
  • 0
    Are you trying a flood fill or flood select type function?

    I started at a spot and checked points next to that one and grew dimensions in the x and y if all points were of same color. I was doing a box fill type function so my goals might be different from yours.
  • 0
    I may not understand the problem but do you want to cluster the image pixels?
  • 0
    Trying to separate connected shapes that sometimes have breaks in them to a certain black and white threshold to clear pixel noise and identify in this case drawn or written characters on a page in preparation for a mnist style training

    In this case in another script than standard English

    Here’s a hint

    All the methods work but in larger noisier images some of them lag a lot

    Some require minor image preparation however
  • 0
    I'm still not sure what the goal is here but it sound like an image segmentation problem. There are countless approaches in that field you can try.
  • 0
Add Comment