Previous | Next --- Slide 13 of 52
Back to Lecture Thumbnails

Why wouldn't we want to concentrate samples where the image is changing the most like at edges?


I still fail to see why we couldn't just sample uniformly, and as long as we take sufficient samples it seems we will still get values from where the integrand is large/small, and we don't need to bother with biasing? Is there an intuitive explanation for this?