[valerio]

Tomas you could indicate through one image to me the set of your scene? Which type of emitter you have used? Which are the parameters of the camera?

Thanks Tomas An!!!!

[Frances]

...Not sure how this speed corrolates to dielectric noise, or situations involving interior architectural scenes. Seperate tests might be needed for those conditions.

Thomas, you are the master of understatement.

[ivox3]

Thomas, ......as soon as I duplicate your exact results, I'll let you know about a job well done. :


Seriously though, ..... expert work.

[markps]

Thomas, or you are a genius or you have a very eccentric mathematical way to talk those charts are insane. Many scientist are not capable of putting out as much empirical detail of things as you did.

[Frances]

Thomas, or you are a genius or you have a very eccentric mathematical way to talk those charts are insane. Many scientist are not capable of putting out as much empirical detail of things as you did.

I think you have to be a scientist to understand it.

[Polyxo]

Very impressive work Thomas! Not that I could say, that I understood everything right away...
Whenever you come up with your private tests, I do good in printing things out and rereading the other day...

cheers, Holger

[DELETED]

DELETED

[michaelplogue]

After reading this a couple of times, I actually understand it now! Thanks Thomas for a unique and concise approach to this subject. You da' man!

[michaelplogue]

Sorry, can someone tell me in one sentence which one is actually faster?

In a nutshell - based upon Thomas' detailed analysis of this scene - V1 is 4 times faster than Beta when achieving the same quality in the noisiest parts of the images.

[markps]

Well, does it mean that the perceivable noise is reduced 4 times faster or that the same amount of "information" on the noisy areas is reached 4 times faster?

Sorry, can someone tell me in one sentence which one is actually faster?

In a nutshell - based upon Thomas' detailed analysis of this scene - V1 is 4 times faster than Beta when achieving the same quality in the noisiest parts of the images.

[Mihai]

Well, does it mean that the perceivable noise is reduced 4 times faster or that the same amount of "information" on the noisy areas is reached 4 times faster?

If the same amount of information were reached the noise would look identical.

[Voidmonster]

For all practical purposes, this test is saying V1 is 4 times faster. That feels right to me based on what I've done.

And I think I can explain why the measured entropy was 20 times faster. It's an artifact of the compression used in the PNG file. Unless I'm mistaken -- and I'm basing my understanding on the method that LZH uses, not the patent-free version in PNG -- the compression algorithm is building a library of similar parts in the image. Entropy is somewhat artificially reduced by the patterning of the noise, which is quite regular.

I think you'd get a more solid entropy number from a straight RLE compression.

Also, Thomas, why didn't you save the images in BMP to begin with and then convert to PNG later? The part where you determined the level of compression that Maxwell is using seemed a bit needless to me.

Minor technical quibbles aside, this is a really fantastic bit of data. It's tests like this that make this forum truly great.

Despite the monumental levels of bickering.

[markps]

Well, does it mean that the perceivable noise is reduced 4 times faster or that the same amount of "information" on the noisy areas is reached 4 times faster?

If the same amount of information were reached the noise would look identical.

Solid color results in highest compression (total image size 700bytes)


Same image, but with random noise added. The randomness factor renders the compression algorithm less effective forcing it to store more data (in this case 41 times more with a total image size of 28828 bytes).

Mihai, That's doesn't seem to be true by Thomas' own sample pictures. Even with the image with it's raised entropy the amount of noise is not what makes the image look worse. What makes the noise looks less perceivable is the diminished variation of tones between nearby pixels. So, even though you might have a raised entropy you could be perceiving less noise and the quality of the image could be better.

We don't know how maxwell's algorithm works. And how much it has changed along the time. Any small change on maxwell's V1 algorithm could be generating different "quality" of noise. The amount of entropy doesn't seem to be a direct measure of the perceived quality of the image because the new (or the old) algorithm could be masking the perceived noise effect.

Does it make any sense or am I totally wrong?

[markps]

Other renders generate noiseless images in seconds. The lack of noise doesn't translate in quality of image. Again, the idea that Maxwell is unbiased is not the same to say that every pixel will have the correct color and tone every time a plixel is placed on the framebuffer. Maybe on different versions of the algorithm it is creating a different quality of noise.

On thing to pay attention is that the Maxwell V1 generates on its early Render time iterations a pattern based noise instead of pseudo random noise. That proves that the way it generates noise has changed. And traces of that noise pattern could have been dragged further in Render time even though it is not visible.

Since the raised entropy is being used to measure the amount of noise. And the amount of noise is not a factor on the quality of the image, (with Maxwell Render) only the amout of perceivable noise is what matters. Any variation on the ratio of the generation of perceivable noise to invisible noise has an adverse effect on the measurements.

[Thomas An.]

Again, many thanks to all here for your comments !