liamcottam / wordcloud Goto Github PK

Right now, I'm brute forcing collision detection by checking each pixel against a rectangle generated by the word which is incredibly slow. Plus, it only computes collision one way and is not even bidirectional pixel-by-pixel.

I've been experimenting with a Bounding Boxes however progress has been slow. Generating a BVH for each word, then offsetting it for every point along the spiral may even worsen performance, especially for low word-count images. For example, a 20 word image on a 400x400 image runs in about 2s. The performance mainly decreases as the board size and word count increases.

If anyone has any further suggestions let me know and I'll look into it.