Previous | Next --- Slide 54 of 79
Back to Lecture Thumbnails

Is there some deterministic way to select the number of samples to use? Also, would we ever vary the number of samples across different parts of the scene, like if a scene is really detailed in one area but solid-colored elsewhere?


Does this mean adding more sample points per pixel or increasing the number of pixels?


Is a high-resolution monitor (1080p->2k->4k->8k) equal to an implicit way of supe-aliasing?


Would this ever pose an efficiency problem? If it increases the asymptotic time complexity, then it would be much slower to increase sampling frequency. But if it doesn't, then what stops us from potentially sampling very very many times per pixel? Why stop at 16 or 32 or 64? How can we determine a sufficient number?


Is there a preferred locations for the super-sampled points inside each pixel?