For small images like the ones you all are testing, enable supersampling (S button). It's in the Output configuration rollout. Disabling it is not an issue for large renders.BbB wrote:Iwan, while you're around, why don't you give us your side of the story on this perplexing blurriness?
Fryrender demo is up!
-
deltaepsylon
- Posts: 417
- Joined: Tue Jan 09, 2007 11:50 pm
Anti-aliasing is accomplished by pixel blurring to minimize the "stair-step" effect caused by highly contrasting adjacent pixels. Aliasing is less of a problem with high res renders because there are more pixels to represent the edges of things. A pixel can only represent 1 color at a time, and when you have very thin objects or high contrast edges, heavy aa filtering can result in loss of detail. Enabling a supersampling algorithm, while more ram intensive, helps to preserve detail in low-res renders (800xN and lower).suvakas wrote:What's the reason?lwan wrote:.
BTW, that bluriness is AA filtering, and is intended !
Actually, filtering is not the term I should have used.
Last edited by Frances on Mon Sep 03, 2007 2:40 pm, edited 1 time in total.
real camera are blurry too. if you shot a 2Mpixel photo with a 8Mpixel camera, full sensor is used and the picture is downsized, so it looks sharp, but the full resolution photo seen at 100% would looks blurry.
actually I think a bit blurry AA looks more photographic and I'm in some way responsible of it, because I asked for it.
actually I think a bit blurry AA looks more photographic and I'm in some way responsible of it, because I asked for it.
I don't know. It's certainly easier to blur a sharp image in postprod than to sharpen a blurry one. With all due respect, it feels a bit like a slightly cheap way to get noise-free images earlier. It has its charm, but it's also a serious limitation for people like me who like to render for print (2400x1800min). Rendering at higher res and downsampling is no option, even with a quad core.
Having said that, most of the stuff I'm seeing on the fry forum looks highly impressive, in particular your MPD system. So hats off.
(PS: you should really implement .obj imports. Can't think of why you wouldn't want to have it...)
Having said that, most of the stuff I'm seeing on the fry forum looks highly impressive, in particular your MPD system. So hats off.
(PS: you should really implement .obj imports. Can't think of why you wouldn't want to have it...)
Granted, I'm a photographer by trade, but:
The reason a real camera's image looks unsharp at 100% is a few different things; mostly it's a result of the AA filter (which is designed to blur anything over the nyquist frequency, to prevent aliasing) over the camera sensor, followed by the Bayer pattern of the sensor itself (which must be demosaiced before viewing), followed again by lens issues, diffraction, etc.
Therefore, the "blurriness" is excusable, out of camera.
As for a rendering engine with no such issues; well, that's for you guys to decide.
My *guess* is that Fryrender/Maxwell are "cheating" slightly, and simply not sampling the *edges* of objects as much as they should. IMHO.
The reason a real camera's image looks unsharp at 100% is a few different things; mostly it's a result of the AA filter (which is designed to blur anything over the nyquist frequency, to prevent aliasing) over the camera sensor, followed by the Bayer pattern of the sensor itself (which must be demosaiced before viewing), followed again by lens issues, diffraction, etc.
Therefore, the "blurriness" is excusable, out of camera.
As for a rendering engine with no such issues; well, that's for you guys to decide.
My *guess* is that Fryrender/Maxwell are "cheating" slightly, and simply not sampling the *edges* of objects as much as they should. IMHO.
That's what the "S" button is for! Fryrender has one. Maxwell does not.zsouthboy wrote: My *guess* is that Fryrender/Maxwell are "cheating" slightly, and simply not sampling the *edges* of objects as much as they should. IMHO.
When you are doing small hobby renders, use the S button. It takes longer and uses more ram, but gives you a sharp image. When you are rendering pro work for printed output, render without supersampling. It is a waste of time and ram for large format output.
Output for print does not have to look as good at full size when viewed on a monitor. The print process downsamples the image for you and it looks good. I do it all the time. But I figure for 8x10 @ 300dpi or 2400x3000 pixels. For 11x17 output, I go for 200 dpi.
not me, they look very blurry. Try the demo sometime, it's really blurry. Now the preview mode, is not as blurry, but also has some fireflies, so you decideOnoSendai wrote:I always thought Maxwell had very sharp images...
Indigo doesn't look blurry without supersampling, just more artifacts, and even with 2x supersampling most of us use, is still pretty sharp. I've tried 3x and 4x supersampling and it doesn't make Indigo blurrier for that matter.
Fry is definitely faster than Indigo, but the renders looks so different with a lot of caustics missing, compared to Indigo in Bidirectional MLT mode that you can go either way. There is a certain flatness to the overall look to me. Speed vs Accuracy?
Fry tends to be a lot brighter as well, but that's probbaly a tonemapping issue, Reinhard in Indigo tends to look grey and dull to me
if anyone wants I can put pics of all 3 renderers side side
Who is online
Users browsing this forum: No registered users and 43 guests