forward raytracing algorithm

Discussion related to the implementation of new features & algorithms to the Core Engine.

Moderators: Dade, jromang, tomb, zcott, coordinators

forward raytracing algorithm

Postby jantschek » Tue Jan 10, 2017 5:55 am

Hello,

may it be possible to integrate a pure forward (lightsource to camera) algorithm into Lux Render?
I think it will speed up rendering time a lot, especially in the case that you're working with tiny light sources like LEDs within huge complex objects.

From my understanding the number of useless traced rays should be reduced in those cases.
Of course such an engine won't work with infinte light sources.
Usually also an integration cone for the camera pixels will be needed to define kind of a tolerance around each image pixel that collects light (and creates a more or less blurry result image).

I'm looking forward to your feedback.

BR
Lux Render Version: 1.7 (in classic mode)
Exporter: Blender
Exporter Version: 2.79
Operating System: Windows 7 X64
jantschek
 
Posts: 187
Joined: Thu Jul 09, 2015 11:02 am

Re: forward raytracing algorithm

Postby Dade » Tue Jan 10, 2017 9:56 am

jantschek wrote:may it be possible to integrate a pure forward (lightsource to camera) algorithm into Lux Render?
I think it will speed up rendering time a lot, especially in the case that you're working with tiny light sources like LEDs within huge complex objects.

From my understanding the number of useless traced rays should be reduced in those cases.
Of course such an engine won't work with infinte light sources.
Usually also an integration cone for the camera pixels will be needed to define kind of a tolerance around each image pixel that collects light (and creates a more or less blurry result image).

I'm looking forward to your feedback.


There is already a light tracer (aka forward ray tracer) in LuxCore. However is there mostly as debug tool for BiDirectional path tracing as BiDir is nothing else than a merge with MIS (Multiple Importance Sampling) of a Light tracer (aka a forward ray tracer) and a Path tracer (aka backward path tracer).

Short version: a light tracer is always useless if you have a bidirectional path tracer available.

P.S. if you know that, in your scene, the light paths are far more important than eye paths, as optimization, you can set long light path max. depth and short max. eye path depth (i.e. shorter are the eye paths and more the output is similar to the one of a light tracer).
User avatar
Dade
Developer
 
Posts: 8318
Joined: Sat Apr 19, 2008 6:04 pm
Location: Italy

Re: forward raytracing algorithm

Postby jantschek » Wed Jan 11, 2017 2:13 am

Hi,

Thanks for your answer. Is there a way to adjust the camera sensitivity?
I mean a kind of a integration cone that is around each camera pixel that defines the allowed directional variation.

BR
Lux Render Version: 1.7 (in classic mode)
Exporter: Blender
Exporter Version: 2.79
Operating System: Windows 7 X64
jantschek
 
Posts: 187
Joined: Thu Jul 09, 2015 11:02 am

Re: forward raytracing algorithm

Postby Dade » Wed Jan 11, 2017 2:40 am

jantschek wrote:Thanks for your answer. Is there a way to adjust the camera sensitivity?
I mean a kind of a integration cone that is around each camera pixel that defines the allowed directional variation.


You seem to describe something similar to the depth of field (i.e. the lens radius parameter).
User avatar
Dade
Developer
 
Posts: 8318
Joined: Sat Apr 19, 2008 6:04 pm
Location: Italy

Re: forward raytracing algorithm

Postby jantschek » Wed Jan 11, 2017 3:03 am

Does the depth of field influence the number of integrated rays?
Lux Render Version: 1.7 (in classic mode)
Exporter: Blender
Exporter Version: 2.79
Operating System: Windows 7 X64
jantschek
 
Posts: 187
Joined: Thu Jul 09, 2015 11:02 am

Re: forward raytracing algorithm

Postby Dade » Wed Jan 11, 2017 6:12 am

jantschek wrote:Does the depth of field influence the number of integrated rays?


The question is not applicable: more samples you render and more rays you trace. As an unbiased rendering (monte carlo integration) it will converge to the correct result, first or later.

P.S. in theory, because in practice, with 32 bit floating points, etc. is a quite complex topic if it will converge to the correct result or not.
User avatar
Dade
Developer
 
Posts: 8318
Joined: Sat Apr 19, 2008 6:04 pm
Location: Italy

Re: forward raytracing algorithm

Postby jantschek » Wed Jan 11, 2017 11:36 am

I don't agree.
All ray tracers that i know, that are capable to perform luminance calculations, use an integration tolerance to speed up the process. As far as i know the chance to have a ray hitting the pixel AND having the correct direction to intersect the observer point is zero (no gathering options available). There must be a tolerance to receive at least something? Am i wrong?
Lux Render Version: 1.7 (in classic mode)
Exporter: Blender
Exporter Version: 2.79
Operating System: Windows 7 X64
jantschek
 
Posts: 187
Joined: Thu Jul 09, 2015 11:02 am

Re: forward raytracing algorithm

Postby jantschek » Tue Jan 17, 2017 6:01 am

Were my thoughts that stupid that there is no answer neccessay anymore?
Lux Render Version: 1.7 (in classic mode)
Exporter: Blender
Exporter Version: 2.79
Operating System: Windows 7 X64
jantschek
 
Posts: 187
Joined: Thu Jul 09, 2015 11:02 am

Re: forward raytracing algorithm

Postby acasta69 » Wed Jan 18, 2017 3:29 am

Your question makes sense... for me at least: my knowledge of ray tracing internals is very limited.
I have a feeling that what you describe is more related to depth of field: gathering rays from different directions should indeed simulate a non-pinhole camera.
Can it be that the probability of having a useful ray is extremely small but not zero, hence the inefficiency of forward algorithm?
These are probably very basic concepts for rendering sw, but I'd be glad if Dade could post a short explanation... :oops:
Windows 10 64 bits, i7-4770 3.4 GHz, RAM 16 GB, GTX 970 4GB v376.54
acasta69
 
Posts: 294
Joined: Fri Dec 20, 2013 3:18 am

Re: forward raytracing algorithm

Postby burnin » Wed Jan 18, 2017 8:06 pm

hi
passing through... and seems as very interesting topic
Are any special papers available for reading to get more acquainted with?
Am still itching for anything that would accelerate the whole process, specifically in caustics department. :oops:
burnin
 
Posts: 252
Joined: Mon May 03, 2010 8:04 pm

Next

Return to Architecture & Design

Who is online

Users browsing this forum: No registered users and 3 guests