Something many in CinC can make use of....

[Didereaux]

An excellent video explaining how to avoid producing technicolor puke when processing your photos.

[DanK]

Thanks for posting, but I think he first of the two parts (using the RGB histogram to avoid overprocessing the photo) is bad advice and shows a misunderstanding of the histogram. His basic argument is that you should look to see if the three colors are not balanced and dial back whichever color is dominant at the high end. In his example, red dominates at high luminance levels. That doesn't necessarily mean anything about overprocessing. A photo with lots of bright reds--for example, many photos of red flowers--will and SHOULD look like that. If the red is clipping--another part of his example--that can be as simple as an exposure problem. This is in fact very common in photos of red flowers and is one reason that several of us here have urged people to display all three colors on the LCD in order to avoid clipping the color channel that is dominant at the bright end.

Here's an example. First, the photo, which has very modest processing, and no saturation adjustment:



Now the histogram, which looks like the one he used to suggest oversaturation (although his also clipped at one point in the video):



This histogram just shows that there are a lot of red pixels at the bright end, and no pixels in the other color channels at the bright end. that's what the image looks like, not a result of processing.

The only reliable way to see whether a photo is overprocessed is to look at it. An RGB histogram is not a reliable substitute.

[Didereaux]

Thanks for posting, but I think he first of the two parts (using the RGB histogram to avoid overprocessing the photo) is bad advice and shows a misunderstanding of the histogram. His basic argument is that you should look to see if the three colors are not balanced and dial back whichever color is dominant at the high end. In his example, red dominates at high luminance levels. That doesn't necessarily mean anything about overprocessing. A photo with lots of bright reds--for example, many photos of red flowers--will and SHOULD look like that. If the red is clipping--another part of his example--that can be as simple as an exposure problem. This is in fact very common in photos of red flowers and is one reason that several of us here have urged people to display all three colors on the LCD in order to avoid clipping the color channel that is dominant at the bright end.

Here's an example. First, the photo, which has very modest processing, and no saturation adjustment:



Now the histogram, which looks like the one he used to suggest oversaturation (although his also clipped at one point in the video):



This histogram just shows that there are a lot of red pixels at the bright end, and no pixels in the other color channels at the bright end. that's what the image looks like, not a result of processing.

The only reliable way to see whether a photo is overprocessed is to look at it. An RGB histogram is not a reliable substitute.

I believe you may have missed what he was doing with that RGB histogram...he was simply watching/adjusting for color CLIPPING at either end. Not the brightness separation between the colors. You also seem unaware that saturation is represented by the vertical component of the histogram NOT the horizontal. Be careful with when you make flat statements that they are in fact correct, otherwise you will mislead those who read your opinions.

[Shadowman]

Thanks for posting, but I think he first of the two parts (using the RGB histogram to avoid overprocessing the photo) is bad advice and shows a misunderstanding of the histogram. His basic argument is that you should look to see if the three colors are not balanced and dial back whichever color is dominant at the high end. In his example, red dominates at high luminance levels. That doesn't necessarily mean anything about overprocessing. A photo with lots of bright reds--for example, many photos of red flowers--will and SHOULD look like that. If the red is clipping--another part of his example--that can be as simple as an exposure problem. This is in fact very common in photos of red flowers and is one reason that several of us here have urged people to display all three colors on the LCD in order to avoid clipping the color channel that is dominant at the bright end.

Here's an example. First, the photo, which has very modest processing, and no saturation adjustment:



Now the histogram, which looks like the one he used to suggest oversaturation (although his also clipped at one point in the video):



This histogram just shows that there are a lot of red pixels at the bright end, and no pixels in the other color channels at the bright end. that's what the image looks like, not a result of processing.

The only reliable way to see whether a photo is overprocessed is to look at it. An RGB histogram is not a reliable substitute.

If you have Lightroom the soft proofing mode might assist as well.

[DanK]

I believe you may have missed what he was doing with that RGB histogram...he was simply watching/adjusting for color CLIPPING at either end. Not the brightness separation between the colors. You also seem unaware that saturation is represented by the vertical component of the histogram NOT the horizontal. Be careful with when you make flat statements that they are in fact correct, otherwise you will mislead those who read your opinions.

Fair enough, he was talking about clipping in several places, but he also refers to using the histogram to figure out which colors are "dominant" and "saturated," neither of which is shown by clipping. I'm on the wrong computer now, but I could show you images similar to the one I posted but in which I accidentally overexposed, leading to clipping of the red channel. What that produced was not more saturation, but a loss of detail in the reds.

The histogram doesn't have saturation on the vertical axis. The vertical axis is the count of pixels at that point on the x-axis, rescaled so that it fits within the vertical space of the display. This is true regardless of whether you are looking at the RGB histograms or the luminance histogram. The X-axis is luminance. So, if you go to the histogram for the image I posted, the bulge of red near the top shows that there are a lot of bright pixels that are red in that image. This is explained in more detail in the 'understanding histograms' tutorials on the tutorials page of this website.

Clipping tells you nothing about saturation. Clipping is simply censoring of the the histogram. Let's take clipping at the top. The maximum value on the x-axis is 255. If the image is overexposed, or if postprocessing pushes the histogram to the point where it clips, all of the pixels in the resulting spike that he refers to are constrained to that maximum value. There is no information left, and they all look exactly alike. So what happens when you clip is you lose detail. In his example, it is the movement of the entire red bulge that contributes to the garish look, not the fact of clipping as such.

[Rebellion]

Very interesting and informative, thanks Dudley

[mastamak]

Very interesting video, thanks Dudley. I am inclined to agree with Dan's criticism of the suggested Photoshop solution but I was quite interested in the video guy's method of tackling the sliders in Lightroom to achieve a realistic post-processing solution. This is a method I managed to stumble across myself by trial and error and it is reassuring to see someone else (an expert?) using the same method.
I am not sure how the use of soft proofing in Lightroom, as suggested by John, would assist as I would have thought that this is a specific tool designed to reveal out of gamut colours for particular printer/paper combination and to allow final adjustments prior to printing. The difference in the soft proofed image for different paper types is quite revealing but I don't think it tells us much about whether the image post processing has been over the top.
Grant

[davidedric]

Not watched the video, but for information Lightroom's soft proofing shows out of gamut colours for both the print and the monitor.

Dave

[mastamak]

Thanks Dave,
I think that clarifies John's advice, above.
Grant

[DanK]

I am not sure how the use of soft proofing in Lightroom, as suggested by John, would assist as I would have thought that this is a specific tool designed to reveal out of gamut colours for particular printer/paper combination and to allow final adjustments prior to printing. The difference in the soft proofed image for different paper types is quite revealing but I don't think it tells us much about whether the image post processing has been over the top.

You're exactly right. Soft proofing tells you whether some colors are out of gamut for a particular combination of printer and color. It doesn't tell you whether an image is overprocessed.

To make this concrete: in general, coated papers have a substantially wider gamut than matte papers. This difference is striking in the case of the two papers I use most often: Moab Exhibition Luster and Red River Polar (matte). It's often the case that I will find no out-of-gamut colors when I softproof for the luster paper, but the exact same image, processed exactly the same way, has out-of-gamut colors for the matte paper.

It's quite easy to create an over-the-top image that won't show out-of-gamut colors for the luster paper, and it is also quite easy to find images that aren't over the top that have out-of-gamut colors for the matte paper.

Whether an image is over the top is a matter of taste, and the only way to tell reliably is to look at it.

[xpatUSA]

Fair enough, he was talking about clipping in several places, but he also refers to using the histogram to figure out which colors are "dominant" and "saturated," neither of which is shown by clipping. I'm on the wrong computer now, but I could show you images similar to the one I posted but in which I accidentally overexposed, leading to clipping of the red channel. What that produced was not more saturation, but a loss of detail in the reds.

The histogram doesn't have saturation on the vertical axis. The vertical axis is the count of pixels at that point on the x-axis, rescaled so that it fits within the vertical space of the display. This is true regardless of whether you are looking at the RGB histograms or the luminance histogram. The X-axis is luminance. So, if you go to the histogram for the image I posted, the bulge of red near the top shows that there are a lot of bright pixels that are red in that image. This is explained in more detail in the 'understanding histograms' tutorials on the tutorials page of this website.

Clipping tells you nothing about saturation. Clipping is simply censoring of the the histogram. Let's take clipping at the top. The maximum value on the x-axis is 255. If the image is overexposed, or if postprocessing pushes the histogram to the point where it clips, all of the pixels in the resulting spike that he refers to are constrained to that maximum value. There is no information left, and they all look exactly alike. So what happens when you clip is you lose detail. In his example, it is the movement of the entire red bulge that contributes to the garish look, not the fact of clipping as such.

Quite so, Dan. In fact, over-saturation (for whatever cause) causes the opposite of clipping (in the HSB/HSV model), namely that colors opposite to the over-saturated color get 'bottomed'. For example, over-saturated yellows would result in a histogram where the blue curve is showing many more counts at the zero level than green or red.

Some flower colors can be seriously bothersome in that respect. Blue flowers, for example. Firstly, blue is the worst possible color for distinguishing luminance differences, i.e. contrast. Secondly, cranking up the saturation on a blue flower to make it more pleasing can bottom both the red and green channels, leaving only the blue to show petal detail with what little contrast it can muster. Remember the formula for luminance Y (the property by which we determine contrast):

Y=R*0.21+G*0.72+B*0.07

With those ratios, blue doesn't give you much luminance content, eh?

Or trying reading text on those really irritating pages that have blue text on a black background.

[DanK]

Here's a concrete example that illustrates why the histogram is the wrong tool for this. I started with the image of the hibiscus I posted above. In the first edit, I just jacked up saturation to create an over-the-top image, pushing it until the reds clipped, just like in the video:



You can see that the color has become unnaturally intense. The image is also beginning to lose detail in the reds, which is because of the clipping.

Then I went back to the original and just boosted exposure using a levels adjustment to get a similar amount of clipping:



It just looks overexposed. The colors are not more intense. There is again some loss of detail because of the clipping.

Below I will post the histograms from the two new images. Those of you who are familiar with photoshop will know which one is which because the levels adjustment creates gaps in the histogram. However, the shapes of the histograms of all three channels are similar, as is the clipping of the reds.

Depending on the mix of colors and brightness, these two simple adjustments won't always produce such similar histograms. One of the three I tried showed the effect that Ted described. However, this one example makes the general point, which is that the patterns described in the video are not a good way to estimate whether your image is overprocessed. It's simpler and more accurate simply to look at the image.