HSV COLOR MAPPING

[R_Cyph]

Hey all,
I was wondering if someone could give me some insight into HSV color mapping. My understanding is that HSV Color mapping is supposed to compress the float lighting into your standard 8bit range.

Question 1: when using HSV color mapping I know HSV is supposed to maitain the saturation and intensity information , but does it also maintain the relationship of light to dark areas? or does it just "clamp" the brightest parts to 1.0?

Question 2: why does nobody seem too use HSV color mapping? Everybody seems to use reinhart and exponential ( which seam to wash out the color if you ask me ) is there some downside I'm not aware? ( besides the obviously loss of float data )

[percydaman]

I use linear pretty much exclusively. Ive found that if you need to use those other color mapping types, then you need to retool your lighting/materials etc. I HAVE tried the reinhard but only when playing with the sun/sky system.

[DaForce]

reinhard at default 1.0/1.0 is linear mapping.

I generally find myself using linear all the time too. that and LWF is all i find myself need usually. The occasional tweak in PS as well.

Hey if it looks good stick with it

[vance3d]

and if you are using render elements, linear is the way to go.

V Miller

[R_Cyph]

and if you are using render elements, linear is the way to go.

V Miller

Why do you need to keep everything in linear space to comp elements?

[R_Cyph]

and if you are using render elements, linear is the way to go.

V Miller

Why do you need to keep everything in linear space to comp elements?

Is there anyway to apply "Color mapping" in post after comping linear elements together?

[joconnell]

Linear is the best way for comping since most compositing functions use regular maths functions like add, subtract and multiply. For those to work correctly across the colour range of an image it has to have a linear transition from light to dark - otherwise you could have something like an add operation not adding enough in the midtones or too much in the shadows etc.

Colour mapping is just applying a curves or levels adjustment, the only difference is if you do it via vray, the anti aliasing can take it into account. Say for example you wanted to apply some really harsh colour mapping on an image where a bright and a dark colour meet, if the colour mapping makes this more contrasted you might get some really nasty edges. If vray is aware that the contrasting will happen, it can take that into account along the edge with its anti-aliasing so you get your harsh colour mapping but no nastiness along the edges.

[studioDIM]

Ah, bloody hell.
Another one of Jo's star post.
I love to get schooled like this

Lele

[joconnell]

I'm paying lele monthly p.r. fees

[R_Cyph]

Linear is the best way for comping since most compositing functions use regular maths functions like add, subtract and multiply. For those to work correctly across the colour range of an image it has to have a linear transition from light to dark - otherwise you could have something like an add operation not adding enough in the midtones or too much in the shadows etc.

Colour mapping is just applying a curves or levels adjustment, the only difference is if you do it via vray, the anti aliasing can take it into account. Say for example you wanted to apply some really harsh colour mapping on an image where a bright and a dark colour meet, if the colour mapping makes this more contrasted you might get some really nasty edges. If vray is aware that the contrasting will happen, it can take that into account along the edge with its anti-aliasing so you get your harsh colour mapping but no nastiness along the edges.

OOOOHHH ok that makes lotsa sense. I was under the misguided impression that COLOR MAPPING was the same as TONE MAPPING.

My understanding on tone mapping is that it takes a float image and mapes those values to standard 0 - 1.0 values , while maintaining the brightness/contrast relationships.

So just to sum up, you guys recommend using linear mapped elements , then comp in post. which brings us to my last question last question...

Is there anyway to TONE MAP in post ?

[joconnell]

Yeah the problem is if the gradual change from light to dark isn't even, then doing a lot of math based post operations will have more or less of an effect across the shadow, midtone and highlight of an image which will look off and it's also less predictable so that's why we like linear.

In terms of colour mapping vs tone mapping you're totally right - it is indeed just compressing the values between 0 and 1, it's just that tone mapping has got some sneakier algorithms built in to it to give a nicer visual result quicker than trying to manipulate curves. You're literally just playing with a float image and taking selective details from different parts of the colour range to make your final image. If you use a curves adjustment on an image you're processing each pixel in the image the exact same way so while you could take a float image and bring back the details in the shadows and the burnt out sky, it's a very simplistic way of treating the image. A more controlled way of doing it would be to take certain details within the picture using either masks or keying and selectively expose them - much like a film telecine or colourist would do. This is a bit more similar to what tone mapping is trying to achieve.

There's a few different methods of tone mapping so I can't give you an absolute explanation that covers all of them but one marked difference between colour mapping and tone mapping is how it decides which areas to bring back and which to leave overexposed. One of them uses contrast between pixels to guess if there any detail worth bringing back - if it's a flat area of colour it's probably a wall or something boring so it'd leave that alone whereas if there's different patterns and bits of contrast it could be something more interesting and bringing back into a useable range - it's be like shooting a hdr image and then keeping the correctly exposed bits from each image and making a single final image from it.

Here's a plugin for photoshop to do it - http://www.hdrsoft.com/ - if you have a look at the three images on the left and the final composite you'll get the idea. Personally I reckon the image looks fake and weird since you could never ever see that with your own eyes - It's why a lot of the interiors you see here look unnatural - the inside is exposed properly but the outside is too - it's something you couldn't do with a single photo and you couldn't see it with your eyes.

Personally I'm sticking with linear as it's the most natural way for me to get realistic images and also if you render to float with vray which would allow you to tone map in post, you might be bringing in aliasing problems and the only way to solve them is to render double size and scale down in post - you'd do your exposure on the double size image and once you scaled it down to regular size you get a crude form of anti aliasing but in most programs you wouldn't get any of the various filters like mitchell, catmull etc (thankfully catmull is pretty much an unsharp mask though).

[R_Cyph]

Well that certainly clears up alot of questions I had about color/tone mapping.

I did a comp test last night with linear mapped elements with a trial copy of combustion. HOLY CRAP!. I'm just loving the control. NO more render times for a minor adjustements. I mean always knew compositing had its benefits but I never really got it to work properly before.

I also had a look at the tone mapping plugin and you're right. Some of those photos do look odd especially where they have gone overboard. However, some of them (IMO) look quite striking. When used properly ( subtle effect ) I definately think tonemapping adds life to images.

I was able to acheive a very similar images to the VFB with HSV color mapping using the float>linear>comped>tonemapped workflow.

ALAS, I couldn't my comped linear elements to match my RGB ouput from the VFB. ( any hints on this ? ) But, even still they were close and I definately like the control I have over everything.

thanks again

[Da_elf]

ill be the moron that says that this post doesnt belong here and needs to be moved. your having a problem with a concept, not a problem with software bugs which is what this section of the forum is designed for

[R_Cyph]

its still a problem nonetheless

[percydaman]

ill be the moron that says that this post doesnt belong here and needs to be moved. your having a problem with a concept, not a problem with software bugs which is what this section of the forum is designed for

wow, I thought I was anal. Ill move it to render theory though.