blowing out a channel?

another weird question I have never asked but always wanted to.

I hear people talk about blowing or clipping a particular colour channel......I'm confused about that. Not sure what it is I'm even trying to ask? I thought if you blew out an area in a pic it's blown out in all colours? As in a blown out sky simply turns white.

I had a look here http://www.cambridgein​colour.com/tutorials/h​istograms2.htm and down the bottom of the page there is a photo of some flowers where the red was clipped and it says this:

Maybe what I am trying to ask is, What makes an image clip one channel but not the others?

pixelbasher wrote in post #16869024
Maybe what I am trying to ask is, What makes an image clip one channel but not the others?

The colour of the object.

In the example you quote the red flower looks red because it absorbs all other colours and reflects red light. So if it's illuminated with bright white light the green and blue light is absorbed and little is reflected into the camera - no clipping. But most of the red light is reflected, so the red is much more intense than the green and blue, so more likely to suffer from clipping.

Clipping of a color channel can show up when that color is particularly bright and/or saturated, and becomes a concern when you want colors to be accurately "rendered" in your presentation and printing of the image. Usually it's not that much of a concern, but if shooting scenes with, like I said, highly bright/saturated colors, it can be a consideration, where you may want to use the RGB histogram and make sure you are shooting Raw to give yourself the most latitude in processing.

In real life and in the LAB color space, something can both be extremely bright or dark and simultaneously very saturated at the same time, but sensors can only see in primary colors and then guess the luminance based on the intensity of adjacent recorded colors. The luminance in an RGB image can never be more intense than a pure primary color, so if a primary color clips, you start losing luminance information, especially if the subject is made up of mostly that color.

In some cases, it's possible to recover luminance using Raw conversion software that can "guess" the appropriate luminance as long as one or two unclipped channels exist and have sufficient information to fill in the blanks with; as far as I know Lightroom since version 4 and Capture One 7 can both do this.

To add to the mix, consider that an image file may have a clipped channel (or channels) because of exposure decisions made a the time of capture, or because of processing following image capture - either by the camera in the conversion to JPEG, or in post during white balancing and raw conversion.

If one shoots in raw, then one can inspect the actual raw histogram of the file without white balance applied to get a sense of whether or not clipping has occurred in the RGB channels. During white balancing, exposure in the R and B channels, typically, is scaled to achieve some neutral specified by the user. This may result in clipping of a particular channel as well. If all of the data are intact in the raw file, but post-capture edits are pushing a channel (or channels) to clipping, then one must consider how to get the desired rendering while managing the clipping. Conversion from a larger working color space to a smaller space can also potentially result in "clipping," or remapping of colors that are contained in the source color space gamut but outside of the destination space gamut. The user can, to a certain degree, select how this remapping takes place by specifying a rendering intent for the conversion from one space to another.

One way to visualize clipping in a single channel is to think of each channel as a black and white image, where black represents the absence of that particular color in the composite image and white represents full saturation of that color in the composite. The attached comparison of a highly blue object shows that, as saturation increases, the dominant color (blue) can clip in the highlight end of the histogram (the channel blows out to white) while the sort of opposite color (the red channel here) clips to black. Green contributes relatively little to the dominant color, so its channel is actually pretty well preserved, even at the extreme end of pushing saturation of the blue car. This channel might be used to attempt to reconstruct the luminance information that gets lost when the other channels clip.

The top image is the straight raw conversion - there is even some shadow clipping in the red channel due to the highly saturation nature of the blue car (orange arrows). The bottom image has had the saturation increased to extreme amounts, and you can see that the channels depict this correspondingly. The red channel has clipped to black in most of the car area, with the red tail lamps on the other cars clipping to white. The blue channel has clipped to white in the car area, etc. When these channels clip, significant detail is lost due to the luminance information getting destroyed. There is no longer detail in the red tail lamps and the contours on the body of the blue car have washed out.

Pushing images like this ON PURPOSE is a good way to condition the individual channels for masks - you can drive the black and white channel to extremes and create a ton of contrast for making a detailed mask without using time consuming magic wand or color range techniques.

Keep in mind that this example was made in the AdobeRGB color space and then converted to sRGB for posting here - when I did that, the extreme blues and reds had to be remapped to fit in sRGB. The change was visible on screen during conversion.

kirk

Man, I gotta get better with using the individual RBG channels, especially for masking purposes.

Kirk, your understanding of the subject is always impressive. One of these days it'll click for me.

Thanks Nathan - the thing is, unless you have some innate ability to visualize an image this way, or have the experience to do so, one of the major shortcomings of raw converters that encourage start-to-finish workflow is that you have no access to the actual channel information, so it can be difficult to analyze and address these kinds of things. If you are getting into digital image processing through digital photography and all you use is Lightroom or a similar raw converter, you may never realize how your image is composed in the RGB world - if you do not have a sense of how colors relate, then you may end up just sliding sliders and getting discouraged with the results.

Photoshop is such a powerful tool, with so many applications, it can be intimidating to try to even begin tackling it - but really it is just a set of tools to apply one's thoughtful experiments with color and tone. Even if you never thought of yourself as an "artist" who paints or draws, learning a little bit of color theory can be really helpful in forming an approach to all sorts of digital image processing tasks. As with all things, there are probably several different ways to perform a task, which can be seen as difficult (which one do I chose?) or really awesome (wow, I can adapt this technique to do what I want!). I tinker an awful lot, and make a lot of mistakes, but also learn a lot in the process. That's the fun part for me.

As an exercise pertinent to the above example - try opening some images in photoshop and guessing what the individual black and white representations of the R, G and B channels look like, without first looking at the Channels tab. Think about not only the tonal value in each channel, but how the same part of the image will look in one channel versus the others (will they all be about equal - a neutral value? will one channel be a lot lighter than the others?). Try looking at the gray versions of R, G and B channels in a single image and guessing what the resulting composite color image would look like - I'm not exactly sure how one would go about extracting the three channels without seeing the color image first, but you get what I'm saying. You could probably make an action that does it.

Maybe I'll post an example here and see what different folks guess!

kirk

Quick quiz!!!

Here are three channels from a color image shot with my crappy phone camera. First, the image should give you context enough to figure out which image corresponds to which channel (R, G or B).

This was taken at my local fire department's demonstration day - it was awesome. They set a car on fire to demonstrate how they work to extinguish the fire safely. They also set the stove on fire and a building they made out of 2x4's and plywood. Pretty cool.

Anyway, see if you can figure out the colors of the following items in the image - this is fairly basic, just try to think about it based on the relative levels of gray in the object across all three channels.

Also, remember that R, G and B and the light primaries:

R+G = Yellow.
G+B = Cyan.
B+R = Magenta.

Okay!

Here are the three channels, in no particular order.

You've no doubt seen the Fried Egg Effect in sunset pictures. Look for an orange ring that forms when the red channel saturates, followed by a marked yellowing, closer to the sun, that sets in when the green channel maxes out.

1: Blue
2: Red
3: Green

A: red
B: yellow
C: orange
D: blue
E: red
F: white
G: black
H: brown
I: cyan

How did I do?
Now I'm gonna try if I can recreate the original image in photoshop.

Yay! I managed to recreate the original, but won't post it yet, in order not to spoil the answers

A) Extinguisher is Red
B) Label is light yellow (there's still a little blue)
C) Washer is almond-y. Not quite white, some blue is missing.
D) Tape is silvery-blue?
E) Flames are red, yellow, and orange. Red is starting to clip.
F) Car is white, but not exposed to white since we see the shadowed side. So, gray.
G) Smoke is dark red.
H) Pavement is dark gray
I) Sky is overcast gray? there's a lot of the red channel still in there.

Thanks for the exercise.

Yes, you can recreate the original! See this super cool old school technique for color photography from black and white film cameras:

http://en.wikipedia.or​g/wiki/Sergey_Prokudin​-Gorsky

and this:

http://www.loc.gov/exh​ibits/empire/making.ht​ml

and this:

http://www.loc.gov/pic​tures/item/prk20010000​01/

Actually recreating the image from the channel components also helps understand how color channels combine to form the composite. @Jefzor - when you get a chance, post a short description of what you did to recreate the image and we can compare it to the original. There are a few different ways I have used before to do this (on the Gorsky image plates, for example).

Here are the black and white plates I used in my first attempt to do such a thing:

You two did very well! Granted there was a lot of context in which to interpret the color, but that is sort of the point. Anyway, great job.

kirkt wrote in post #16872371
@Jefzor - when you get a chance, post a short description of what you did to recreate the image and we can compare it to the original. There are a few different ways I have used before to do this (on the Gorsky image plates, for example).

_______________

Here are the black and white plates I used in my first attempt to do such a thing:

Can you guess which "channel" is which?

Here is the unedited result:

1: Blue
2: Green
3: Red

would have been harder to guess without the final result


As for my method, I simply pasted the images into their corresponding color channel in photoshop (see attachment). Note that the arrows become grey because they have the same brightness in each source image.