3

I just got an Epson V600 and have been trying to learn some of the features, such as the Histogram Adjustment feature of the included processing software. The interface looks like this:

enter image description here

What are the general principles for setting the arrows?

I notice that often the default settings have the black/white arrows pointing into the histogram, instead of bracketing it. Why is this? I would think that the user should put the arrows just outside of the histogram (as shown) to capture the full range of color, or no?

What does the X-axis of the histogram represent?

1

Yes, clipping certainly will change colors, making them lighter (sometimes desirable), and worse, for example clipping red more than blue can add color cast. That is an issue beyond just the loss of any highlight detail.

However, clipping (modest or sometimes more) can often greatly help many B&W images, increasing contrast, bringing them to life, making them sparkle. Blacker blacks and whiter whites is B&W contrast. It's a standard practice for B&W work. Caution, excessive contrast is not good for most color work.

Some things to know:

One of Ansel Adams basic tenets (in B&W work) is that there should be some degree of pure black and of pure white, for "strength and conviction". His work has few exceptions about that. He did it with darkroom manipulation, but clipping both ends can do it too.

X value is the [0..255] color shade of the tone. All histograms are 256 wide.

Y value is the count of pixels with that tone. It is scaled to be relative count, the maximum height will always nearly reach full scale.

So histogram shows how pixel tones are distributed.

In Photoshop Levels, and in ACR Exposure, holding the Windows ALT key while touching the slider with mouse, will show WHICH PIXELS are being clipped at that setting (to decide if clippings lost detail is important).

Note that (for color images) a one-color gray histogram (called Luminance) is quite different than a three-color three-channel histogram. The single gray histogram cannot show color clipping.
See http://www.scantips.com/lights/histograms.html

  • "All histograms are 256 wide." This is not correct. While it is often the case, particularly for apps that work in 8-bit per channel color, many apps will use histograms of different widths, particularly if they work in 16 or 32-bit per channel color. – user1118321 Jun 5 '17 at 2:00
  • A 16-bit histogram would be 65536 pixels wide, which is a bit too much for my screen. :) For 16 bit data, Photoshop will still show the 8-bit histogram then, 256 pixels wide. Measure it. – WayneF Jun 5 '17 at 2:12
  • Sure maybe PS will, but other apps may choose something different, such as showing it at 655 pixels wide, or compressing the histogram to whatever size the user makes the window. There's nothing inherent about making a histogram 256 pixels wide. A histogram is a general statistical tool not specific to image editing. But even within image editing it can be any size. – user1118321 Jun 5 '17 at 2:45
  • If you are not just making things up, then name your example that we can see. – WayneF Jun 5 '17 at 3:02
  • Well, there's HDRHistogram. This article shows a histogram with only 10 buckets. This article from Wolfram shows how you can plot an HDR image's histogram in a very wide range (in this case about 750 pixels wide). – user1118321 Jun 5 '17 at 3:15
1

Yes, your assumption is generally the way I handle levels. It will, of course, depend on the image you are scanning. If you're scanning an intentionally low-contrast image - perhaps a light colored object in fog - then you would not want to do it that way as you'd be introducing contrast that wasn't intended. But for typical content, you are correct.

Typically, the x-axis represents either the luminance or channel brightness of pixels, and the y-axis is the number of pixels in the image at that luminance. So the left-most arrow is the black point. Wherever you put that in the histogram, that luminance will be mapped to 0. Everything below it will be clipped to 0. The right-most arrow is white point. It gets mapped to 255 and everything above it gets clipped to 255. The middle arrow is gamma and affects the mid tones in the image.

I notice that the output black point is set to 18 rather than 0. I normally leave the output set to 0 and 255 unless there's a valid reason not to. (For some video formats 16-240 is preferred.)

1

It depends on the bit depth you are scanning with, the output format, and your subsequent workflow.

If you're scanning at 16 bit, but producing a JPEG which you don't want to edit further, you have to set the end points and the gamma slider so that a pleasing image is produced. That can sometimes be inside the image data, to clip specular highlights and true blacks. It may even be useful to set the colour channels independently to correct colour shifts.

If you want to edit the image further, but choose an output format with lower bit depth than you scanned with, you'd better move in the black and white points until you touch the tails of the histogram, because you'll want to make optimal use of the bit depth available.

If you want to scan to a "raw" image in full resolution for further processing, you can leave the sliders at 0/255, of course.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.