Area 51 » Technology

The Computer Graphics site didn't have enough activity during the beta and has been closed. You can download the data dump of all questions here.

220

followers

51 Example Questions

active newest votes

up vote 23 down vote

added by Dan the Man Nov 14 '13 at 16:02

Reminder: Each question needs only 10 upvotes. Beyond that, the votes are essentially wasted and could be better spent in some other question needing more votes. – Victor Stafusa Feb 12 '14 at 7:28

@Victor Well said. – Dan the Man Feb 21 '14 at 17:32

All the people who have already voted for this question are free to remove their vote so that they can vote for a worthy question which still needs to be brought up to 10. Provided this one doesn't drop below 10 it will still count towards the target. – githubphagocyte Mar 16 '14 at 11:45

It's also worth taking a glance through to see the 5 questions you have upvoted. If they have exceeded 10 votes since you voted for them you can remove your vote and use it lower down. It's worth checking for this frequently as new votes keep coming in... – githubphagocyte Mar 20 '14 at 15:04

up vote 18 down vote

added by MickLH Nov 27 '13 at 20:56

you must send the url for your code (in github) or at least say what are you doing (opengl or directX, using omni or other ligths, etc). – Pedro Polonia Feb 19 '14 at 0:48

This site is still a proposal my friend! Perhaps I will start a community wiki with that if it passes! :) – MickLH Feb 21 '14 at 18:43

up vote 16 down vote

added by Dan the Man Nov 18 '13 at 3:54

up vote 16 down vote

added by CharlesL Nov 27 '13 at 19:51

up vote 15 down vote

added by CharlesL Nov 18 '13 at 13:56

up vote 14 down vote

added by Dan the Man Nov 14 '13 at 16:00

up vote 14 down vote

added by Dan the Man Nov 18 '13 at 4:45

up vote 14 down vote

added by zedutchgandalf Nov 28 '13 at 10:49

This isn't strictly true, but is only a trend in the current GPU hardware. The situation can be inverse on some completely viable rendering hardware. – MickLH Nov 30 '13 at 8:34

up vote 13 down vote

added by noob Nov 28 '13 at 19:56

up vote 11 down vote

added by Dan the Man Nov 14 '13 at 16:38

up vote 10 down vote

added by CharlesL Nov 18 '13 at 17:30

up vote 10 down vote

added by zedutchgandalf Nov 28 '13 at 10:59

up vote 10 down vote

added by AbigailBuccaneer Nov 28 '13 at 11:58

up vote 10 down vote

added by Trinidad Nov 28 '13 at 16:22

Great question. – Dan the Man Nov 28 '13 at 18:54

up vote 10 down vote

added by Trinidad Nov 28 '13 at 16:23

up vote 10 down vote

added by heishe Dec 14 '13 at 22:31

Explanation: Glossy reflections aren't trivially implementable when using image-based light sources, because glossy reflections contain information from many different directions, and a simple texture lookup only samples one direction. Efficiently doing this is not trivial. Example answer would be to have a mip-map chain of the used IBL probe containing the different reflection values based on glossiness of the object, and then to interpolate between the different mip map levels somehow (perhaps just linear interpolation looks good enough). – heishe Jan 24 '14 at 23:04

Weighted average of the surrounding points in the lighting image, using a Gaussian distribution? A small enough sample over a wide enough area may be sufficient for a visually acceptable reflection. – githubphagocyte Mar 14 '14 at 22:23

up vote 10 down vote

added by heishe Dec 28 '13 at 21:05

Explanation: z-prepass is a technique used to save fill-rate performance by drawing a depth-only rendering pass to populate the z-buffer. this (is supposed, at least,) to avoid the performance penalty from over-draw, for example when rendering the g-buffer pass in deferred shading or just normal rendering of meshes in forward-rendering. – heishe Jan 24 '14 at 23:07

up vote 10 down vote

added by Caleb Jan 21 '14 at 16:34

up vote 10 down vote

added by Caleb Jan 21 '14 at 16:39

up vote 10 down vote

added by heishe Jan 24 '14 at 23:10

Looking for historical significance or algorithmic justification, to get a feel for the way things are done. Either way plenty of scope for citations and examples, and likely to give fresh insight for a large number of readers. – githubphagocyte Mar 13 '14 at 20:31

up vote 10 down vote

added by Brian Rushton Feb 5 '14 at 16:41

I don't think this is a place for questions about animation and art related techniques. This is more about technical computer graphics programming and techniques. – Dan the Man Feb 21 '14 at 17:36

Personally I would be happy to see animation questions if they concern the updating of an image and techniques to ensure consistency between frames - that seems nearer to the programming side of things. I agree that art questions are off topic - I wouldn't want to see "help! my animated cat looks more like a dog..." – githubphagocyte Mar 31 '14 at 20:31

up vote 10 down vote

added by Brian Rushton Feb 5 '14 at 16:42

up vote 10 down vote

added by Brian Rushton Feb 5 '14 at 16:46

up vote 10 down vote

added by apiad Feb 19 '14 at 16:11

Explanation: In classic Monte Carlo rendering algorithms (Path Tracing, MLT, Bidirectional PT), a fixed number of rays is shoot through each pixel. All optimizations are done in the object space (importance sampling, etc.), i.e, for rays that belong to the same pixel. However, we could also optimize in the screen space, spending less rays in "easier" pixels (which converge faster) and more rays in more problematic areas (such as caustics). This question is about exposing existing techniques for determining the optimal number of rays to dedicate to each pixel, to maximize overall image quality. – apiad Feb 19 '14 at 16:16

This is the kind of question I would really like to see. If there's anyone who doesn't consider it worthy of an upvote I'd be fascinated to know why. – githubphagocyte Mar 13 '14 at 20:29

up vote 10 down vote

added by githubphagocyte Mar 13 '14 at 20:12

up vote 10 down vote

added by githubphagocyte Mar 13 '14 at 20:19

up vote 10 down vote

added by githubphagocyte Mar 13 '14 at 20:23

up vote 10 down vote

added by githubphagocyte Mar 13 '14 at 20:26

up vote 10 down vote

added by System Down Mar 18 '14 at 18:37

up vote 10 down vote

added by System Down Mar 18 '14 at 18:52

up vote 10 down vote

added by System Down Mar 18 '14 at 18:52

up vote 10 down vote

added by zedutchgandalf Mar 18 '14 at 23:23

up vote 10 down vote

added by Gassa Mar 20 '14 at 14:53

up vote 10 down vote

added by System Down Apr 5 '14 at 5:32

up vote 10 down vote

added by cloudraven Apr 6 '14 at 6:30

up vote 9 down vote

added by Mokosha Nov 28 '13 at 15:15

up vote 9 down vote

added by heishe Dec 1 '13 at 10:11

up vote 9 down vote

added by Mokosha Dec 9 '13 at 0:06

up vote 9 down vote

added by cww Dec 17 '13 at 10:45

up vote 9 down vote

added by heishe Dec 28 '13 at 21:12

Explanation: Traversing a tree (such as an octree) in shaders can be problematic, due to traversal usually including many branches [bad on many graphics hardware architectures] and shaders having no stack available can cause problems also. – heishe Dec 28 '13 at 21:15