Simulating in Photoshop the usage of filters in B/W photography
1) What is the easiest way to do it?
Of course, the input (scanned image) should be the color.
Then I want to see the effects of different BW filters (e.g. yellow #8, minus-blue filter.. etc) for the situation as if I shot that image on BW film using this or that film.
2) The spectral characteristics of brand name filters are well known (Kodak, Nikon...). Is there a way to simulate it in Photoshop?
3) Almost identical question: is there a more or less precise method to simulate the changes in color temperature? ('Precise' means 'can be measured'; just moving the slider in Hue adjustment window is not enought..)
In addition, if we had a way to simulate the char.curve of a given pair "BW film + its development mode" we have an excellent learning tool for BW photography!
#1. "RE: Simulating in Photoshop the usage of filters in B/W photography" | In response to Reply # 0AlanC Basic MemberWed 19-Jul-00 03:46 PM
To answer your first point, the technique I've used for changing the relative amounts of colours that go to make up a black and white image is to use the channel mixer: Image menu, Adjust, Channel Mixer..., then click the Monochrome option on the dialog.
The snag with this technique is that you're working with the Red / Green / Blue source channels, and I'm not sure if you'd be able to accurately imitate particular filters. The question would be whether the filters have a linear effect on the film. Anyone know?
#2. "RE: Simulating in Photoshop the usage of filters in B/W photography" | In response to Reply # 0Wed 19-Jul-00 06:29 PM
on question 1:
the things you ask for can be done with some knowledge of the color balance, hue/sat, and channel curves, but it can be easier done with the Filter Factory.
you can input functions to change colors with 8 possible sliders, for R,G,B and A. for a minus-blue filter, the result would for instance be:
r: i - (ctl(0) * b)/150
g: i - (ctl(0) * b)/150
b: i - (ctl(0) * b)/150
the function "i - (ctl(0) * b/150)" means the corresponding grayscale value of the color (according to the standard color-to-grey conversion) minus the value of blue in this pixel.
to make the filter variable, CTL(0) is added, which is the first usable slider, and the quotent is devided by 150, so's to get a certain under/over filtering. is the controller is at 150, the image will be (intensity - blue component).
Yellow #8 is even easier. just add a set value to R and G:
(this value might not belong to yellow #8 btw, this is just an illustration of the simplicity involved =)
for more flexibilty, you can add control sliders for a more variable addition.
on question 2:
If you know what the combinations do, you can easily write a quick plugin yourself, or spend a bit more time in the curve editing.
on question 3:
yes. I wrote a very rough plugin that does "cooler" and "warmer", but it's not much trouble to write a reliable (I didn't use charts as standards for my plugin) plugin that has a color change corresponding to the selected difference in Kelvins with a slider. It's all a matter of how accurate you want to make it, but if you invest a little time, you can achieve things in Filter Factory in more time than the normal editing would take, but you can save it as plugin, and just run it very fast after the first time.
I've written a small set of plugins that I needed a while ago which are hosted here at nikonians, so you can see what you can do with FF if you feel like giving it a shot. All the filters in it didn't take more than an hour to program. (and most of it was clearing typos =)
#3. "Real light and Color models (was: Simulating in Photoshop the usage of filters in B/W photography)" | In response to Reply # 2Mon 24-Jul-00 06:52 AM
Alan and Mike, thank you.
Now I want to go into details deeper and therefore some more questions:
1) Mike, you mentioned "standard color-to-grey conversion".
I remember my early days of PC learning, there was a VGA BIOS function performimg the similar conversion according the formula
w = K1*r + K2*g + K3*b
where w -- brightness of the resulting gray tone ("whiteness")
r,g,b -- values of reg, green, blue in RGB model
K1, K2, K3 -- constant coefficients (I don't remember exact values).
Important, that it is a linear function.
The 1st question is: is this formula correct for RGB -> Grayscale recalculation in currently used color models (e.g. in Photoshop)? Does anybody know the coefficients?
Alan, your question probably means either similar _linear_ function can be considered as a formula describing the _real_ BW film response. (If YES then applying a filter is equivalent of changing the coefficients' value, isn't it?). Looking at the film sensitivity graphs published by manufactures it is hard to believe. The question rather should be: if we take the linear approach how big will be the error? --- I cannot tell right now: the mentioned graphs show dependence of sensitivity as a function of real _wavelength_, not of RGB-values. Therefore my second question:
2) does anybody know the relationship between real wavelenght (lambda)and RGB-values?
The real spectrum looks like Hue range in Hue/Sat/Br model, but Hue range ends at magenta (after deep blues); it is not the case in the real spectrum.
3) By the way, does anybody know the conversion formulas between different color models used in Photoshop:
RGB <-> HSB
HSB <-> Lab ?
Again too much questions in the single message...
#4. "RE: Real light and Color models (was: Simulating in Photoshop the usage of filters in B/W photography)" | In response to Reply # 3Mon 24-Jul-00 07:10 AM
1) it certainly is. the current RGB-gray conversion formula is:
Intensity = (76*Red + 150*Green + 29*Blue) / 256
2) I'm working on a plugin that creates a BnW height map according to wavelengths. you'll have to bear with me, because it's not easy to convert RGB values to a corresponding wavelength value. Or rather, it is if you've got one given color, but to generate an algorithm that does it with any RGB value is a bit tricky. Hue and wavelength don't match evenly. while Hue is linear, wavelength color-shiting isn't.
3)HSB should be easy to convert, as it stands for Hue-Saturation-Intensity. I don't know about lab colors...
#5. "RE: Real light and Color models (was: Simulating in Photoshop the usage of filters in B/W photography)" | In response to Reply # 4Mon 24-Jul-00 08:51 AM
look at http://www.efg2.com/Lab/ScienceAndEngineering/Spectra.htm
there is an downloadable spectra.zip (with exe) and some related information. Though I didn't find a good theory till the moment.
#6. "RE: Real light and Color models (was: Simulating in Photoshop the usage of filters in B/W photography)" | In response to Reply # 5Mon 24-Jul-00 01:53 PM
uh-oh.. this graph is.. um.. complicated to do in short-code programming used for making PS filters =)
It might take time, but I'm sure that this is very usefull for creating a reliable plugin.
#7. "RE: Real light and Color models (was: Simulating in Photoshop the usage of filters in B/W photography)" | In response to Reply # 6Tue 08-Aug-00 11:00 AM
The novice is trying to sum up.
The most important question is how the real spectrum correlates with any color model say with RGB.
How can we emulate the color? From point of view of physics --- a very respectable point of view --- the closer are the spectra the better is the match. Thus if I need a discrete spectrum representation I should take several wavelengths spread across the whole visible spectrum, call it “coordinates” in this color space and assign appropriate values in such a way that the sum of all this narrow band functions were as close as possible to the spectrum to be emulated. – A well-known procedure in mathematics (step-function approximation). And the more elementary functions I take (the more is the number of color coordinates), the closer is the approximation. For example “red, orange, yellow, green, blue” (5-D color space) sounds good. The “dark red, red, orange, yellow, yellow-green, green, light blue, blue, dark blue” (9-D color space) sounds even better.
But when I run Photoshop (or just turn the TV on) I’m pretty happy with all the range of colors represented there using RGB model, i.e. only 3-D color space! How is it possible? Any approximation of the spectrum with only 3 narrow band functions looks very poor, if you graph such an approximation!
The answer is probably in physiology of the human vision. The eye asks for much less than the spectra matching. The eye maps the wide variety of different spectra to single hue (as it is percepted by brains), so there are, roughly speaking, much more shapes of spectrum than the number of perceptable colors. Essentially is the conclusion that we lose information when we switch from spectrum measuring to the color vision (and therefore to the appropriate color models). It means that having a given RGB-representation of a color we are not able to recover the ‘true’ spectrum.
But the property of a filter is known in terms of spectrum!
Now, to simulate the effect of a filter in B/W photography we strictly speaking have to:
I. Shot the subject several times in different monochrome light (lighted with red only, with orange only… etc – the more the better). Each image should be considered as B/W.
II. Combine those images with weights proportional to the filter’s transmission at each color used in step 1.
III. If we take in account the different spectral sensitivity of the film, repeat the step 2 with the film spectral sensitivity curve
Since the white light is the sum of red, orange etc the steps I-II can be replaced:
1) Shot the subject once, in white light only
2) Then decompose the color image into several physically “true” colors (red, orange etc) --- NOT into RGB parts! – and apply filter’s transmission curve
3) Combine resulting set of monochrome images.
4) Apply the step III if desired.
Sounds tremendous, but real problem is the step 2 only: how to decompose the image in the software which uses come predefined color models only! A kind of spectrum analyzing device seems to be unavoidable…
Am I wrong? Are there less pessimistic ideas?
#8. "RE: Real light and Color models (was: Simulating in Photoshop the usage of filters in B/W photography)" | In response to Reply # 7Tue 08-Aug-00 01:00 PM
Well , you lost me. What is it that you specifically want these filter mimics to do? while a filter is of course, manifested in real life, and manipulates the visual spectrum, the result is a static black and white image with very easily definable characteristic that can be approximated with a greyscales very well.
The effects should be of course concidered through a Hue and Intensity point of view, but the result will be black and white.
The main problem is the fact that if you want to apply a filter TO a black and white image, there is only your mind's eye to tell you what has roughly what color in real life. the computer doesn't. a certain grey value might correspond to blue, or to dark yellow. There's no way to say with certainty.
Applying a filter to a color image, to MAKE it black and white on the other hand doesn't require exact knowledge of the spectrum of visible light at all. or perhaps just a little. all you need to know is what the original image looks like, what that same image looks like with a certain filter hooked to the lens, and then emulate it's effects. knowledge of the spectrum helps, but it's not strictly needed.
The great thing about 16 million RGB values is the fact that we can't see much more than 6 million of those values as seperate colors. N-d color space is very nice for real life color synthesis, but on any device, you can reduce the entire image to 3d color space, and you can make it look convincing. not because it's approximated, but because with three color compenents, you can create more colors than you can distinguish.
The idea of matching RGB values to the visual spectrum is a bit crippled due to the fact that the visual spectrum doesn't end, but simply "fades", varying per person. the RGB values that correspond to them therefor aren't "closed" at either end. if you look at that "Spectra" graph, there are no (R255,G0,B255) values, while there are clearly very purple colors too. the Spectrum isn't exactly a 3d color space, but rather a 3d color space PAIRED with an intensity value.
There is no real need to decompose an image in as many colors as possible, apply the filter curve and reassemble them, but rather you can apply an alternate curve (the unified curve for the color scheme you're working in) on the image. A filtercurve that can be applied to a 9d color space can be rewritten to fit a 3d color space with the exact same effect. that's what maths were invented for =)
#9. "RE: Real light and Color models (was: Simulating in Photoshop the usage of filters in B/W photography)" | In response to Reply # 8timmy Basic MemberSat 06-Jan-01 10:09 AM
WOW ! I'm in awe of the responses ! Exceptionally informative (and accurate) information here ! Must be a whole-bunch of Photo Scientists. For "just quick and dirty simulations", I simply "select all" of the R, G or B channel and copy it. Then, open a new window and paste it. I've never been too concerned about matching what would have been "captured" using B&W film and a filter (and all the associated light source and film development variables).
Can I present a "reality check" here? Once we are working on an image, most folks will be using 8-bit information (24 bit RGB or possibly 32 bit CMYK). To perform "accurate" color metric transforms, more data would be desirable. Unless you can "get" the raw-scan data (12 bit would be nice!), I doubt you could do a "first class" job. On the other hand, "how accurate does it need to be?". If we consider 8 bit as representing status A reflection density times 100 (0-255 = 0.0- 2.55), that's beyond any reflection print I've ever heard of.
#10. "RE: Real light and Color models (was: Simulating in Photoshop the usage of filters in B/W photography)" | In response to Reply # 9wardseward Basic MemberSun 21-Jan-01 12:17 PM
I've seen a product advertised in the back of Digital Photo magazines that cames to simulate all kinds of lens filters and more. I've have yet to download the demo and try it, but feel free to and let us all now how it work. =)
It's called nikColor Pro. http://www.tech-nik.com/
They have some there software listed at the site also, but I haven't had the time to read any of it. Sorry I can't give more input on this one.
#11. "RE: Real light and Color models (was: Simulating in Photoshop the usage of filters in B/W photography)" | In response to Reply # 10jnscbl Basic MemberSun 21-Jan-01 01:08 PM
I always wondered what MENSA members talk about when they get together
I believe it is said that a physical filter of a particular color can only enhance that color on film,(by suppressing other colors) and that it cannot add it. I have noticed in my own tiny experimentation with photo software that you likewise cannot boost a color that isn't already there.
I do not possess this higher knowledge of relationships between reflected color and additive color, but I can't help but wonder: wouldn't the algorithmic duplication of a glass filter, of a particular color, be a simple matter of suppressing the remaining colors?---scott
"I am always doing that which I cannot do, in order that I may learn how to do it."
#12. "RE: Real light and Color models (was: Simulating in Photoshop the usage of filters in B/W photography)" | In response to Reply # 11BJNicholls Charter MemberSun 21-Jan-01 10:51 PM
From what I recall, black and white film gained the name "panchromatic" when chemists engineered an emulsion that was sensitive to a full spectrum of light. You're right that color filters selectively block spectral components by absorbing them, thereby changing the tonality of the black and white image.
When you do a scan of a color slide or negative, you won't be getting quite the full spectral range that came in through the lens, but it should be close enough to work with. Say you want to apply a red filter: I'd create a red layer in Photoshop set to "overlay", then convert to greyscale. You could also try doing the same thing with the layer set to "multiply" and once you've converted to grey, go into levels and reset your white level setting so your highlights aren't grey.
Keep in mind that scanners are subject to blue channel noise, so it's quite possible to make this noise more visible when playing with colors and converting to grey. I often go to the channels palette and click on red, green and blue to see which looks best as a grey level image and then I copy the channel an create a new image by pasting the channel into a new document.
#13. "RE: Real light and Color models (was: Simulating in Pho" | In response to Reply # 12dlegros Registered since 02nd Mar 2002Mon 11-Mar-02 03:09 AM
This all makes interesting (but to me quite confusing) reading. The Nik-Pro option looks expensive ($200 for the full set).
I would strongly recommend checking out www.silveroxide.com
The produce a range of plug-ins that are supposed to simulate real-world film charcterstics. I have used their HP5 filter for a couple of years. They now produced enhanced versions that include Red (25A), Yellow (8) and Green (11) filters during the conversion.
The plug-ins are $55 per film type with free demos.
There are only two ways to deal with beauraucrats...
Stealth and sudden violence.