Meaning of vertical lines on a histogram

<p >I understand what the distribution of vertical lines along the horizontal axis on a histogram means. I do not understand what the meaning is when those vertical lines reach the top of the graph. The various definitions say something like they represent the # of pixels @ that exposure level. Well, OK, does that mean that those are over exposed pixels? Obviously not. From what I can tell it makes no difference whenever the # of pixels reach or 'flat-top' the top line.</p>

<p >So what if anything does it mean? If it means nothing, why not?</p>

<p >Thank you, Jay</p>

<p>It's my understanding that the height is arbitrary. If it were proportional, the Y axis would have to be all the pixels in the camera, while the X axis is the dynamic range. A proportional graph would be either ridiculously high, or would place most of the information in a normal image along the very bottom, or both. So the ceiling is lowered, and when more pixels hold a particular value than the graph can show, the graph is clipped. </p>

<p>The histogram vertical lines do not give you any information that can help you make the photograph look better. They are a representation of a number of pixels at given tone value in relation to the rest of the histogram.</p>

<p>I get a lot of spiky histograms with a lot of vertical lines some of which form a boot shape where most of the data is bunched up on the left which shows a rather high contrast and dark preview. I can change the shape, height and smoothness of the histogram by reducing contrast, making the black point lighter and applying an isolated curve in the deep shadows and now the histogram has tapered end points and one big hump in the middle with no spikes or vertical lines.</p>

<p>I've redistributed/remapped the data represented in the histogram to make the image appear as the eyes would see of the original scene.</p>

<p>Don't be concerned about vertical lines in the histogram especially shooting Raw. Be concerned about the overall shape. You can change the shape of the histogram editing jpegs, it's just that you may not have as much shadow/highlight detail because the incamera processor through out a lot of data converting the 12-14 bit data to an 8 bit jpeg.</p>

<p>Vertical lines in the histogram endpoints will show how poorly the data was redistributed/remapped by the incamera processor from the linear sensor data by examining the level of taper in the histogram endpoints in relation to spikes that point to gaps in these areas.</p>

<p>A spike along a histogram indicates an extreme number of pixels at that tone, for whatever reason. Most typically it's due to image degradation, some sort of posterizing going on. Say for example a blended exposure image, where you'll notice unusual halos at some tonal transition zone. The tone of those numerous halos will likely cause a spike.</p>

<p >Matthew, Tim, & Mendel</p>

<p >Thank you for your answers. This is the 1st, for what ever reasons, I have heard any explanations of the vertical lines. I obviously have a lot to learn about digital photography.</p>

<p >Mendel, could U refer me to a photo or two that have an example of what your are referring to.</p>

<p >Tim,I shoot raw only. I don't see the advantage of space savings as worth loosing most post-processing power. It takes me months to fill a 16gig card w/ 16mb images. From my film background, what it sounds like what U are doing is changing the gamma locally @ different places to better match what the eye sees. Now to learn how to do that in Elements or Lightroom.</p>

<p >Jay</p>

<p>Jay,</p>

<p>A given height doesn't represent a fixed number of pixels, but it's not arbitrary either.</p>

<p>Consider what would happen if the vertical lines represented a the number of pixels. Imagine two images, one with all pixels within a narrow range of luminance, and the other with a fairly uniform distribution across levels of luminance. If the height were simply a count of pixels, one or the other of these would be useless. if the scale were such that you could see variation in the the latter, the former would go off the top of the scale.</p>

<p>So I think the histogram is rescaled to fit, given the nature of the image. I have never seen what algorithms the various manufacturers use, but it is close enough to think of the lines as representing relative frequency. </p>

The vertical height represents how full that channel are for that value of color in the color space you have your camera,

raw processor, or post-processor set to.

<blockquote>

<p>From my film background, what it sounds like what U are doing is changing the gamma locally @ different places to better match what the eye sees. Now to learn how to do that in Elements or Lightroom.</p>

</blockquote>

<p>If you have Lightroom 4 & up with PV2012, it's much easier than in previous PV2010 & PV2003 because each slider tool isolates particular tonal zones for editing while maintaining definition especially in shadows shooting Raw.</p>

<p>With regard to using the histogram vs relying solely on a dark preview in ACR/LR that at times hides a lot of data especially in the shadows viewing at "Fit In Window" zoom, I didn't even follow my own advise.</p>

<p>As an example below I've reworked this image at least three times through three PV versions and failed to zoom in on detail that comprised 1/10th the image area which is the dark trees. I'ld always assumed if I lightened it too much it would just show posterized black splotches so I didn't push it and let the image remain dark in that area.</p>

<p>Then one day I looked at the triangular shape (tapered sides) of the histogram section (see red arrow) representing those dark trees and adjusted the Shadow slider in LR4 to 100 viewed at 100% and was blown away on how clean and detailed it was. That histogram represents ProPhotoRGB output which I have to set in Proof Preview.</p>

<p>Live and learn the hard way I guess.</p><div></div>

<blockquote>

<p ><a name="00dTHG"></a><a href="/photodb/user?user_id=5880808">Jay Drew</a>, Sep 02, 2015; 03:54 p.m.</p>

</blockquote>

 

<blockquote>

<p>Mendel, could U refer me to a photo or two that have an example of what your are referring to.</p>

</blockquote>

<p>This is the sort of thing I had in mind. It's a HDR image taken with an IPhone, and at the peripherals of the highlights you can see an odd putty-like zone. Funny enough, the histogram shows little or no spiking, so maybe my example is poor, lol.</p>

<div></div>

<p>When you change saturation, contrast etc, the vertical lines and colors that they represent, change in size. Are there any hints how these might ne used to improve post processing? </p>

<p>The only caveat about saturation/contrast increase induced vertical lines is whether it clips on the end points.</p>

<p>At least that's what I had to conclude a while back on the Adobe forums when a poster asked how to prevent the blue channel from clipping in the shadows because the stock agency he was submitting this particular image wouldn't accept any images with clipped channels. I told him to isolate a node near the zero point on the blue channel in point curves and push the zero position slightly up so the histogram showed the spike moving slightly to the right and no longer clipping.</p>

<p>This slight edit didn't change the look of the image but technically qualified it as acceptable by the stock agency.</p>

<p>Other than that saturation/contrast levels need to be judged by the preview instead of the histogram. And my 56 year old eyes seem to be having problems in that area so I take a break and allow my eyes to adjust and come back to get a fresh look. Most of the time I do alright, but in certain high contrast, wide dynamic range scenes such as sunsets I really go overboard and I adjust accordingly. </p>

 

<p>Good point about preview. I do trust my eyes for the most part. Working with film that's scanned is very challenging to get to colors "normal". I try to adjust to reach a point that looks pleasing to my eyes rather than looking at charts, graphs, etc. I do use the histogram initially to get with range. Then adjust by eye afterwards. I figure if it looks OK to me, then it should be OK with others. I also use a calibrated monitor so I know the colors are at least presenting themselves correctly. Being 70, I understand your point about age. Getting old sucks. </p>

<p>And judging saturation in Bridge thumbnails and Main Preview of Raw files is difficult because they don't show reduced (or intended) saturation seen in 100% zoom previews caused by the Detail panel settings. My thumbnails and Main always appear to be somewhat over saturated.</p>

<p>They're also generated in sRGB space. I hunted down thumbnail/Main previews in Bridge's Application Support directory, copied one to desktop and opened in Photoshop and got the warning that they are tagged in sRGB.</p>

<p>LR's Library view is pretty much the same but I'm not sure if their previews are generated through sRGB.</p>

<p>The vertical scale in the histogram is scaled automatically to occupy the full height of the available space for the histogram.</p>

<p>In statistics a histogram is usually called "frequency distribution". The vertical scale shows the frequency of each binned range of values. The sum of the height of all the vertical lines will always be the number of pixels in the image (or 1 if you prefer to view it as a relative frequency distribution).<br>

Hope this was of some help. If not, print it out, curl it up and throw it in the dustbin.</p>

<p>Regards,</p>

<p >Thanks to all who contributed. I learned a great deal. Sorry to be so late to reply. I was waiting for further comments, got bissey w/ other things,in particular how to Ittl & CLS w/ a Nikon D7000 & SB-800. I have questions about that which I will ask in an appropriate forum. </p>

<p > Thanks again for all the info about histograms & post processing, I learned far more than I bargained for: Jay</p>