Why Blender Cycles puts some white dots in the image?


(Michael George) #1

untitled

Actually, this is the first time to use Cycles and I can see that there are a lot of white points or dots in the rendered image.

I wonder why this happens? and how can I remove them and get a clear image?


(ajarosz) #2

t’s not only cycles thing. It is just a noise and this type of noise is called often here fireflies. When you use caustics in reflection/refraction there is a lot of noise. So, more samples and turn off caustics. If you don’t want to turn off caustics you needd to use like a lot more samples or filter glass/use some clamp.


(Emir Sinan Gürlek) #3

Take a look at this: 7 Ways to Get Rid of Fireflies Once and For All


(sundialsvc4) #4

I’ve pondered for a long time: “what is it about the Cycles algorithm that causes fireflies?” Did the algorithm simply fail to find a light-source for a particular pixel, so it “left it ‘blank?’” If so, then why are the fireflies always small? And always white, not black?

I know that we find various ways to prevent them and/or to compensate for them when they occur, but why, technically, does this [set of] algorithm[s] produce them at all?


(kesonmis) #5

Simplified example: you have a surface that is somewhat rough but reflective, and a small but strong light source. Render with one sample. Each pixel of that surface either reflects that light or not. As ray direction is random, most pixels are black because that one sample ray does not hit light source. Some are very bright because they do hit the light. These look like fireflies. The more samples you add, the more neighbouring pixel values converge. But if light paths that cause bright lighting are very scarse (caustics, light through small holes etc), fireflies need a lot of samples to average out.


(sundialsvc4) #6

A fine, workable explanation. Thank you.

And if I rigged up several light sources of different colors, I’d see colored fireflies of various hues?


(kesonmis) #7

Yes.

Another simplified example of sampling would be something like this. Take four pixels, each sample returns either 0 or 1 until 15 samples is reached:
00000 00110 00111
01001 10000 10010
00000 00000 11111
11111 00000 00000

Samples are averaged as they are collected, but it should illustrate that different distribution of bright and dark samples affects the current brightness of pixels by a large degree. After one sample, three pixels are black, and one is white. After two samples, two are black, one is white, one is gray and so on. After 15 samples they are all the same gray because when averaged together they produce the same result, but the difference in order causes fireflies until enough samples are gathered and pixel values converge.

As each sample has three color channels, hitting colored lights, metallic reflections etc also modifty resulting sample colors and contribution, so you get colored fireflies.