POV-Ray : Newsgroups : povray.pov4.discussion.general : Ideas. New f_distribution inbuilt function. f_hash usage. (povr branch). Server Time
19 Oct 2021 18:26:50 EDT (-0400)
  Ideas. New f_distribution inbuilt function. f_hash usage. (povr branch). (Message 1 to 2 of 2)  
From: William F Pokorny
Subject: Ideas. New f_distribution inbuilt function. f_hash usage. (povr branch).
Date: 15 Aug 2021 10:42:45
Message: <611927e5$1@news.povray.org>
A central intent for f_hash is a method by which we can create pseudo 
random streams from geometric, functional, color or whatever "seeds" - 
on the fly.

A sub aim is to use these f_hash "streams" as input to some of the 
common statistical distributions to create patterns - on the fly.

Elsewhere, I'd mentioned a plan to perhaps replace/extend the 
distributions in rand.inc with those shipped with c++11 onward. This 
idea hasn't panned out well - for a few reasons. Thinking rand.inc will 
stick around as is for a while.

What I've decided to do in the near term is add an f_distribution() 
function from which we can 'run' various distributions. I've only got a 
handful going, but it looks promising.

This means of implementation also allows alternative modes which are 
likely not useful in a purely statistical applications. For example, the 
geometric distribution supports a continuous value return mode in 
addition to the usual discrete value one.

An image is attached showing four of the implemented distributions of 
f_distribution as patterns. Using a cylindrical gradient into f_hash() 
as the initial seed(s) to the stream(s) required.

Bill P.


Post a reply to this message


Attachments:
Download 'distributionstory.jpg' (500 KB)

Preview of image 'distributionstory.jpg'
distributionstory.jpg


 

From: William F Pokorny
Subject: Re: Ideas. New f_distribution inbuilt function. f_hash usage. (povrbranch).
Date: 15 Aug 2021 10:51:13
Message: <611929e1$1@news.povray.org>
On 8/15/21 10:42 AM, William F Pokorny wrote:
> 
> An image is attached showing four of the implemented distributions of 
> f_distribution as patterns. Using a cylindrical gradient into f_hash() 
> as the initial seed(s) to the stream(s) required.
> 
...
The image attached shows the binomial distribution in the top row and 
the geometric in the bottom row. The left column represents the 
distributions without AA, but otherwise matching the previously posted 
image for those two distributions.

In the right column showing the effects of AA - because it's interesting 
to look at AA this way :-).

Normal (povr am2 random jitter) AA with the binomial in the upper right.

Povr's big jitter (well outside pixel, per pixel random sampling) with 
the geometric distribution in the lower left. With this one there are 
'harmonics' starting to show up as bumps between the usual discrete values.

Bill P.


Post a reply to this message


Attachments:
Download 'binomialgeometrichist.jpg' (18 KB)

Preview of image 'binomialgeometrichist.jpg'
binomialgeometrichist.jpg


 

Copyright 2003-2021 Persistence of Vision Raytracer Pty. Ltd.