Understanding Histograms

[Rex]

Hi there,

Have been lurking for some time and as you will see, this is my first posting. So, thought I would make it a biggie!

I scan my mf film and am getting pretty good results. Usually, the histogram looks good the initial adjustment. Frequently, I will adjust the image in LAB, which can send the histogram off the scale if I am not careful.

However, sometimes, the histogram is clipped at one end or the other but when I look at the RGB histograms, only one channel is actually clipped.

What exactly does this imply? I understand if the highlights or shadows are clipped in all channels, but how can one channel be clipped? Surely too much blue (clipped at the highlight end) is blue?

Grateful for some clarification,

Many thanks,

Rex

[mistermonday]

Re: Understanding Histograms

Rex, welcome to RP!
The histogram is merely a graphical representation of the pixel values and how they are distributed across an image. The histogram can have any shape or distribution and that does not mean there is anything necessarily wrong with the image. For example, if you took a photo of a bright yellow moon againt a pitch black sky, in the blue channel histogram, you would see the entire graph clustered over to the far left side of the scale because there are no bright blue pixels.
Clipping can occur if you make a levels or curve adjustment and you lose information. For example if, in a curve adjustment, you set the hight point to be 230, any pixels that had a value of greater than 230 will be gone (or set to the max 255). This will occur if you adjust any channel individually or if you adjust the composite channel, then it will occur in all of the channels at once.
If the histogram is bunched up against the highlight or shadow end, it could mean that the image data is being appropriately represented OR it could mean that someone or something (like a digital camera) processed the image and in doing so already clipped the shadows or highlights. In that case the image has been damaged and may or may not be easily recoverale, at least to its optimum state.
Regards, Murray

[Rex]

Re: Understanding Histograms

Murray,

Thanks very much for your reply and explanation. I know that some images can look OK even though the histogram is bunched or even clipped.

One of the difficulties I have is that most of my output is for photo-libraries. I don't shoot digitally but scan my transparencies. One of the first things an editor looks at (if they plan on accepting the image) is the histogram. Any clipping and generally, the image is out. Most libraries want the histogram to be between about 3 and 252.

Rex

[mistermonday]

Re: Understanding Histograms

Rex, Then I would recommend when you scan you photo (or slide or negative), turn OFF all of the Auto options on the scanner. Do not let the scanner adjust the contrast, nor the saturation. No descreen, No sharpening, No color adjustment of any kind. Take the raw scan and import it into photoshop and make your own level adjustments if you need them. If the scanner is of good quality and has a reasonable dynamic range, the file you get should be 100% acceptable in terms of having no clipping.
Regards, Murray

[Rex]

Re: Understanding Histograms

Murray,

Have nothing on when scanning. I have a Minolta Dimage Multi-Pro and get good quality from it.

The imported raw data is then run through PS and saved as a tiff. Usually, the histogram looks OK at this point but sometimes, I find I get better colour if I correct in LAB. Having gotten the image to look the way I like, I then convert back to RGB and that is when the histogram can look a 'little' wonky.

Perhaps I need to be a little less enthusiasm in LAB. But that is not really the issue; I was just wondering what (or why) one RGB channel can be clipped and what that actually means.

Rex

[mistermonday]

Re: Understanding Histograms

LAB is a very powerful colorspace, but also a very dangerous one. Very small moves in the A or B channels will shift colors dramatically. The color space has a gamut that can not be matched by any other color space. LAB is good for boosting weak colors or separatng colors which are very close together. Frankly, given you have a Dimage scan output, there should be no reason for you to use LAB unless you have a special need. If you are using it just to sharpen the L and not messing with the other two channels then there should be no effect on the R,G,orB when you convert back.
Regards, Murray

[Rex]

Re: Understanding Histograms

Yep, I'm pretty much a beginner with LAB and am trying to understand the Dan Magalis book. Certainly use it for sharpening.

Regarding the scanner output, don't know what is depends upon, but sometimes everything looks good on the screen, other times, quite a lot of tweaking is necessary. Guess it has more to do with film stock/exposure/subject etc. I do not have the scanner profiled; just using the Minolta default but that seems to be pretty good.

Rex

[mistermonday]

Re: Understanding Histograms

If you negs or slides are older than 10 yrs, the emulsions have likely degraded and the film may be faded or have developed a color cast. If you film is in excellent condition but the photography was not that great (over / underexposed) then yes they will probably need some work. Your scanner is most likely giving you an accurate representation of what it is seeing.
Regards, Murray

[Rex]

Re: Understanding Histograms

Some are older than 10 years, most are Velvia but certainly the shooting conditions contribute a lot, as does the memory or how it should look.

Rex