Simulating Proper color

[Aaron2004]

Hello,

 

I am working with a client and trying to pick proper colors for townhomes. In photoshop, I recreated the color the best I could by using siding samples at my desk. However, after I take them, I don't know how I can guarantee that the colors are correct. They aren't the same as in photoshop...and I don't expect them to be, after being in VRAY with a sun, etc.

 

The whole process seems very unscientific. How does everybody get around this? I'm not talking Linear Work Flow here. I'm talking that one little tweak could make the siding look totally incorrect, without me even knowing it.

 

Thanks for any insight,

 

Aaron

[mbowers]

Never never never guarantee colors in a 3d rendering. During the construction process you can specifiy to have the actual paint on a peice of the actual siding to show your client. This is the only way to get the true color you want. Hope this helps.

[Sawyer]

I have had meetings with architects and talked about this. I took a swatch and photographed it inside the buillding then outside brought that up on a screen and showed them the true colors or the colors as seen from my camera to the screen and how much the color varies with conditions. Then most people either print the work themselves or take it to a print shop so it looks different than the screen did. Then I take the images into ps and tweak them so they look better.

Now there was a good example of masked selections and color adjustment in one of Brians articles. That may give you more controll.

Sometimes I think clients think too hard over this more so than if it was a real photo. Maybe you need to talk about adding a color board to the display.

[Eric]

This is something I've been working on for a while. I've created a Photoshop file that is my standard material swatch reference file. We use many standard colors in our office for most of our projects because they're palettes that work well together. So I have, for example, a Sandlewood Beige swatch (among many others) that I've adjusted the color in photoshop to match the output from my printer and the actual material swatch.

 

Now that I have my little swatch PSD file with tons of swatches in it (and I add to with each new color on a new project) that are calibrated to MY printer, I at least have a starting point for "accurate" colors.

 

Take the RGB color value from the photoshop PSD file for that particular material swatch - and plug the RGB value into your 3D material's color. Once your image is rendered out, drop it into photoshop and make all your post processing adjustments. Use your swatch PSD to sample the color for your desired material, and create a new layer in your rendered image PSD. Fill a little square of color on top of your paint color on your rendering. Now, create an adjustment layer under your swatch, and adjust the rendered color until it blends into your swatch. Squint your eyes to blur the image and adjust until you can't tell any noticeable difference.

 

After you've made these adjustments, your colors should be very accurate with your (hopefully) accurate swatch PSD file that was already calibrated with the printer's output.

 

Obviously, colors will vary from printer to printer, and under different lighting, but as long as you're calibrating everything under your own working conditions, and adjusted your rendered output similar to what I've mentioned above - it's probably about as close as you can get and anything more needs to be covered with actual finish boards.

[Eric]

Additionally, you can make sure your "sun" and "sky" are both set to white, instead of yellowish and bluish. Both of those colored lights will obviously have great impact on your rendered output (though they're obviously intended for photo-real results), but when you're fighting over colors with clients, this is an easy step to knock out some of the initial output problems.

[Aaron2004]

Thanks for all the feedback guys...I have to take some time to soak this in, but I'll be sure to get back to you and let you know what's up.

 

Thanks!

 

Aaron

[STRAT]

never hastled me before. my renders are impressions. not exact colour guides to specify by.

 

if my client is unhappy with a particular colour he'll advise me until he's happy. all 'eyed up' of course.

[Crazy Homeless Guy]

vlado's post on matching a color exactly is relevant if there is one color that needs to be spot on.

 

other than that, i would recommend working in linear color space, with a calibrated monitor, and a spectrometer or manufacturer specified RGB values. that should get your color close, with the only variance being lighting condition.

 

 

 

I had to deal with this recently, so I thought it may be useful.

 

The problem:

 

We want to obtain an exact RGB value for some (small) surface in our image. Let's say the RGB value is xr, xg, xb which range from 0 to 255.

 

The solution:

 

1) Render the scene in any way you want with any settings you want, but make sure you use Linear color mapping with 1.0 for both Bright and Dark multipliers and the "Clamp output" option is off, and you are using the V-Ray camera "Exposure" setting. It would also be handy if you render to the V-Ray VFB.

 

2) Measure the RGB value at the pixel you need to match. If the values are, let's say (pr, pg, pb) again from 0 to 255, adjust the V-Ray camera "White balance" to be (pr*255/xr, pg*255/xg, pb*255/xb) and multiply the camera ISO setting by (xr+xg+xb)/(pr+pg+pb).

 

Note that this works only for Linear color mapping. In principle it can be done for any color mapping, but the numbers are hard to find by hand.

 

Best regards,

Vlado

 

 

.

[Crazy Homeless Guy]

and a spectrometer

 

let me clarify, with a spectrometer, or at least the one we have, you can set it on the material, and sample the color. it has problems with reflective materials, but for matte materials, it works great. we can get our paint chips within plus or minus 1 of the actual RGB values.

 

..also, you can sample RGB values for paints from sites like this,

 

http://www.colorcharts.org/ccorg/

 

but keep in mind, if you are working with a LWF, a gamma correct of 2.2 will need to be applied to the color.

[AJLynn]

That looks like a good system. Still, I'd be very careful about any promises involving the colors coming out of any computer graphics, since monitors and printers vary so much. It's an incredible hassle even to match your own printer to your monitor, let alone Adobe vs sRGB vs whatever gamut and your client's printer - and you know that if they print it or color copy it and the building doesn't match their copy they'll be mad at you.

 

I've worked on jobs where we presented black and white graphics, and separate color boards with actual samples, for this reason, and that was corporate work. Since most of my current firm's work is for public institutions we can't specify products by vendor so we can even show a client one color chip and they get something a bit different because the contractor bought a different company's product. We usually put Pantone colors in our spec since it's a much more standard standard than most other things out there.

[Crazy Homeless Guy]

That looks like a good system. Still, I'd be very careful about any promises involving the colors coming out of any computer graphics, since monitors and printers vary so much. It's an incredible hassle even to match your own printer to your monitor, let alone Adobe vs sRGB vs whatever gamut and your client's printer - and you know that if they print it or color copy it and the building doesn't match their copy they'll be mad at you.

 

I've worked on jobs where we presented black and white graphics, and separate color boards with actual samples, for this reason, and that was corporate work. Since most of my current firm's work is for public institutions we can't specify products by vendor so we can even show a client one color chip and they get something a bit different because the contractor bought a different company's product. We usually put Pantone colors in our spec since it's a much more standard standard than most other things out there.

 

i guess the goal has to be to get the colors correct regarding the actual color they are in real life, including realistic values for trees, asphalt, sidewalk, etc.. this way when the image is printed, or viewed on another monitor, all the colors in the image are equally off. after it leaves your office, it is out of your hands, and there is nothing you can do about it.

 

i always found it astonishing how bad color is on cheap tv's. go to the electronics store, and look at a row of tv's. the color is different on each one. i wonder if people with high end home entertainment systems have a spectrometer to calibrate their setup. if not, they should.

[Jeff Mottle]

This is actually an interesting subject to me, as lately I've been getting back into my research and studies on color management and device profiling. I'm going to be speaking about this at a conference in Spain called Mundo Digitales this summer (http://www.mundosdigitales.org/)

 

There are two routes to going about this. One is as you are all suggesting, including vlado, and to adjust the image after it's been rendered to make the color the same RBG values as the original sample swatch. This first route may please the client, which in the end is all that matters. But...The second is the correct way. What you are doing while I guess it appeases the client, is simply wrong and does not reflect reality whatsoever.

 

The technical term for what is happening is called color constancy, but it's almost the opposite of this that is happening, presumably becuase the client knows it's a fabricated image and not a real photograph.

 

Color constancy is sometimes called "discounting the illuminant". It's something built into our nervous system that allows the brain to be fooled into ignoring changes in lighting conditions when viewing an object. A good example of this is color bleed in a white room. In fact this is a really good example to show your clients to explain why their color does not match in the rendering. Render a plain white room and put some highly saturated objects in the room and render it with GI. Naturally the walls will render with tones of those saturated objects from the color bleed. Now take the RGB value on one of the "white" surfaces with the color bleed. Obviously it's not 255/255/255. Print whatever color is it on a seperate sheet. Now ask the client to tell you what color the walls are. Unless they are trying to overthink it, they should tell you white. Now show them the actuall RGB print out of one of those surfaces. You get the point.

 

Had the client taken a photo of the finished building, regardless of the time of day and lighting conditions, in most cases the actual RGB value is not going to match, but if they have the actual material sample in front of them would say that is the material that is on the building.

 

Going back a step now in trying to ensure that the color of the material IS a match in the computer before it's affected by any illuminants, there are a few things that should happen. One is that you should be working in a color managed environment. That means your monitor is calibrated to a known white point, black point and gamma. Then develop a profile for that monitor. Any monitor calibration product will do this. You also need to have your printer profiled (printer, ink/paper combinations). Then you need to ensure photoshop, if that is where you do your printing from, is also set up to work with color management. All of this is a much bigger subject, but serves only to illustrate you need to be in a managed environment before you make the match.

 

Now what you can do is visually match your physical sample to a swatch on screen, although this is the least accurate way to do it. Also given that your monitor is calibrated to a known white point and illuminance, you need to ensure your physical swatch is also viewed in that same situation. Normally this is done with a lightbox where you have a light (illuminant) that is a 6500K or 5000K bulb with a high CRI (Color Rendering Index) and an adjustable brightness control. Assuming your office lights are not overpowering it, you can also have your own makeshift desk lamp with one of these same bulbs. I bought one from Home Depot that is a 5000K lamp with a 0.88 CRI (1 being a perfect light source that can reproduce the colors of various objects being lit by the source).

 

The most accurate way to mesure the actual value of the swatch is to use a spectrophotometer which is a device that can measure the spectral properties of a material. It does this my shooting light at the material and measuring how much light at each wavelength is reflected back. The higher end calibration packages use spectrophotometers rather than colorimters, which will allow you to measure both reflective (paper etc) and emmissive (monitors) surfaces. It will measure the values in reflection at varying wavelenth segments, but most of these packages come with a a program that will give you the LAB or RGB values. You can then enter this exact value into your color swatches.

 

End of the day if making the image wrong is what the client wants, so be it, but at 3pm in the afternoon or 6 pm at sunset, that is not the color that is really being reflected to the eye off of the sample. At least not that matches the illumination of the scene as a whole. What they want is their material to be rendered as though it were in a controlled lightbooth and the rest of the scene rendered to look like with soft "magic" lighting.

 

Anyway, hope this was somewhat informative.

[Jeff Mottle]

It's an incredible hassle even to match your own printer to your monitor, let alone Adobe vs sRGB vs whatever gamut and your client's printer - and you know that if they print it or color copy it and the building doesn't match their copy they'll be mad at you.

 

Over the next 6 months I hope to curb this hassle. I want to create a series of articles about color management as it pertains to arch viz and put it in more laymans terms that allow anyone in our industry to quickly get up and running in a color managed environment.

[Crazy Homeless Guy]

The higher end calibration packages use spectrophotometers rather than colorimters, which will allow you to measure both reflective (paper etc) and emmissive (monitors) surfaces.

 

informative stuff. i thought we were doing well, but i guess we still have a long way to go.

[hockley91]

I get really frustrated with the whole color problem....

 

I've already calibrated my monitor with "Adobe Gamma", but the printer I use, is really awful....the prints are sooooo bad.

 

I think in the end it really depends on what client you have. I've worked with some people who don't really care about the 'exact' color and are quite happy with the renderings, while some clients are really nitpicky about the exactness of the color quality.

 

I guess we all just need to be prepared for it.

[jackb602]

Over the next 6 months I hope to curb this hassle. I want to create a series of articles about color management as it pertains to arch viz and put it in more laymans terms that allow anyone in our industry to quickly get up and running in a color managed environment.

 

That would be incredibly helpful Jeff. In the meantime, can you recommend a good book or web site that covers this?

 

Jack

[Jeff Mottle]

That would be incredibly helpful Jeff. In the meantime, can you recommend a good book or web site that covers this?

 

Jack

 

Yup, check out my post here: http://www.cgarchitect.com/vb/13893-monitor-printer-calibration.html I have a link to one of the best books there.

[sonofalion]

Over the next 6 months I hope to curb this hassle. I want to create a series of articles about color management as it pertains to arch viz and put it in more laymans terms that allow anyone in our industry to quickly get up and running in a color managed environment.

 

Jeff, I 'll be really looking forward to this, have been messing up with my monitor and printer and never got where i wanted. So, let us know when you're done, I'll be gnawing my fingers off in anticipation in the meantime...

 

Paris