New generation rendering algorithm


#1

In point based rendering techniques, more super samples in a pixel area give lower variance. In other words, we expect improved quality of the image with super sampling. But we can not always estimate the required number of samples, for example in the case that there is an extremely bright extremely small region at a pixel.In this research, we introduce a new concept for rendering called a Micro Point which was inspired by from quantum mechanics. Micro Points handle high frequency components in a scene correctly without super sampling. Today it takes much computational cost to render stars in the night sky, hair, or objects whose area is smaller than a pixel by existing methods. By using Micro Points, the calculation cost will be improved tremendously for such cases.

Please check it.
http://micropoint.jimdo.com/paper/


#2

Hope to see more practical images in gallery. :slight_smile:


#3

As stated in the paper, MicroPoints seems very similar in concept to micropolygons. Could you point out some of the major differences between the two techniques?


#4

Algorithm is similar. But the concept is complete different. Micro Polygon is point sampling.
So it need to the super-sampling for anti-aliasing. Micro Point is not use super-sampling. Please read paper for detail.


#5
 Hi, noouch,
  There are two differences between micropolygon and MicroPoint.
  
  First,
  Micropolygons are diced by screen coordinate and natural coordinate.
  Thus, when dicing into micropolygons, it does not consider other properties such that variation of normal vector, uv coordinate, and so on.
  MicroPoints are diced by screen coordinate and these several properies to reduce the variation.
  
  Second,
  MicroPoint normalize a high-frequency component into a low-frequency component by coverage-transparency law.

#6

Iigura,

Your paper looks very interesting.

Just a few questions as it relates to stochasitc sampling.

Will this work for motion blur, fine displacement, and depth of field as well? If so, how does the speed compare to prman?

Will this sampling method work with raytracers like vray or will it only work with your internal renderer?

Renderman has had the corner on decent fast displacement for years, will this allow other renderers to finally assume parity in that respect? If so, are you considering licensing your algorithm so it can be affordably included in vray, maxwell, etc?


#7

>Iigura,

Reply from tacwon.
> Your paper looks very interesting.
Thank you.

> Just a few questions as it relates to stochasitc sampling.

> Will this work for motion blur, fine displacement, and depth of field as well?
Yes.
Today’s MicroPoint version. FireBall ver. 2.
Supported MicroPoint based shadow (Point Light).
Next version Thunderbolt ver.1(Out of core implement) will be supported in these features.
Development schedule: http://micropoint.jimdo.com/schedule/

> If so, how does the speed compare to prman?
Prman is point-sampling and MicroPoint is area-sampling.
Fundamental algorithm is completely different.
So, in this comparison is no means.
But our final goal is real-time caustics rendering implementation.

> Will this sampling method work with raytracers like vray or will it only work with your internal renderer?
Only work internal.

> Renderman has had the corner on decent fast displacement for years, will this allow other renderers to finally assume parity in that respect? If so, are you considering licensing your algorithm so it can be affordably included in vray, maxwell, etc?

Our project purpose is research. It is not comercial purpose.
If Thunderbolt Ver.2(for GI)implement finished,we will be opened in this source code.


#8

ITE2012 paper presentation document(PDF) uploaded yesterday.
This document wrote in Japanese, but include many illustrations.
Maybe it is supported to recognized the detail of Micro Point paper.
Please check it.
http://micropoint.jimdo.com/paper/

English version please waiting.


#9

Now, English version upload.
Please check it.


#10

This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.