histogram

[fjerins]

Just started with digital SLR D70 from shooting film.
when taking pictures, what is the best setting for getting good pictures? I've been using "P" and "A" for general picture taking.
Also, I'm using a SB800. When using the flash, what is the best setting for over all flash photography? I have been using iTTl Bal
flash. This is all new to me and changing from film to DSLR is a challenge.
Any suggestions for educational books on the subject?

[Iwerk]

Your histogram data should be pushed as far right as possible without blowing the highlights.

I shoot manual with my D70s.

The beautiful thing about your camera is that you can take a picture, download that picture into the "Nikon Capture" program, adjust the gamma "custom tonal curve", so that the photo looks exactly how you like it, then upload the software corrections into your camera. Then, take the exact same photo again and the picture comes out of the camera, exactly the way that you like to see it.

There are lots of custom tonal curves for your camera on the internet. Try some of them. The default tonal curve that is already in your camera is pretty dull and lifeless. Nikon puts it that way to prevent you from blowing the highlights.

Most folks don't understand exactly what gamma is or why it is so important. I'll try to explain. I'll use a picture tube as an example but the same principle applies to sensors. No two picture tubes are the same. They all have small differences. If I were to apply 1 volt to each picture tube, some tubes would illuminate brighter than the others. So, we make a gamma table to figure out what voltages are required to make each tube illuminate to the same intensity. The gamma corrections are designed to make all of the picture tubes look exactly the same even though they are not.

The Nikon D70 and D70s cameras have the ability to make gamma corrections right inside the camera. The Nikon Capture software is used to make these corrections. If your images always appear dark, dull and lifeless, then a gamma correction to your camera will make these images look bright and beautiful. I almost sold my D70s when I first bought it because I couldn't understand how my Sony Mavica camera could make my pictures look so good and my Nikon pictures looked so dull. The "custom tonal curves" is what made me love this camera so much. It is a very important feature of this camera.

[Iwerk]

I almost forgot to explain why it is so important to get the histogram data as far to the right as possible. It involves binary data which all computers work on.

The way that the histogram information is stored is like this:
- the dark parts of the image are on the left of the histogram.
- most of your colors are in the middle of the histogram.
- the really bright parts of the image are on the right.

The histogram data starts from left to right. As you progress from left to right you have more binary information available.
As an example: 0110 1001 0101 1100
- The left-most group of binary digits can range from 0 to 15 in our numbering system. So, the dark parts of the picture have a limited number of intensities.
- The 2nd group of binary digits can range from 16 to 127.
- The 3rd group of binary digits can range from 128 to 1023
- etc, etc, etc. I hope that you see what I am getting at.

As the histogram information is presented from left to right, you get more levels of color changes as things get brighter. ( to the right of the histogram )

If your histogram was not pushed to the right, then you would be losing out on some really big levels of color. It is these different levels of color that make your photos look even better.

You can directly compare this to your computer monitor. If your computer monitor could only display 16 colors, then it would look pretty boring. But if you had 16 million colors, then things are starting to look really good.

[rpnix]

OK; I'm confused....

There's no more or less "room" for data at the left end of the histogram than at the right... The histogram shows the amount of pixels at each light level. So let's think about this a bit...

A pixel is represented as an RGB value (Red, Green, Blue). Each of these values is one byte long, giving a fixed range of values from 0 (all bits off) to 255 (all bits on). These three bytes together give us the "24 bit color model" we've come to know and love.

If all three bytes in the pixel are near the zero end, then the pixel is dark (000000 hex being black). If they're near the 255 value, the high end, then the pixel is light (FFFFFF hex being white). Colors are made by mixing the three primary colors, so a medium high value for R and G, with a lower value for B, will give you a yellow. (I think; It's early here...) It'd look something like EDA522 hex.

That being said, there's a continuous scale of brightness across the range, with no bias to the high or low end.

What you'd really like for your histogram would be a traditional school exam bell curve (which I usually blew for everyone else, but that's another story...) Very few pixels at the low end, and very few pixels at the high end, with the majority grouped largely in the middle. Things you don't want to see are a large spike at either end. If at the left end, then the photo will be dark, and you'll be losing details in the shadows (although this isn't really a problem if you're taking a picture of a black wall.... every rule has exceptions). If the spike is at the right end, then you'll have very bright, or washed out spots in your photo, again losing detail.

You also don't want your bell curve skewed to the left or right of center in most cases. If it is, you're losing contrast. If you take a picture like this into an editor (PS, GIMP, PaintShop Pro...) and drag the curve to the left and right ends, you'll be amazed at how much better the picture will look, because you'll have spread the curve out to use all the pixel values, instead of being bunched at one end.

Remember high school and shoot for the bell curve.

[Iwerk]

Sorry for not responding earlier but I have found the website where I got this information from. Maybe they can explain it better.

www.luminous-landscape.com/tutorials/expose-right.shtml

[Mark Barendt]

I'm with rpnix.

I think the luminous-landscape guy is all wet, in a practical sense.

The histogram only tells you how many pixels are at a given level of brightness and you only lose info when the graph starts "climbing the walls" at either end. When it is climbing the wall it is recording those pixels as absolute black or white and no detail.

The other thing to remember is that all this varies with your white balance setting because the RGB channels change their relationships to each other (right or left) with a change of WB.

If the WB is "goofy" the histogram on the camera will falsely skew to right or left.

Histograms are great tools but they also don't tell the whole story.

If you are spot metering in camera on someones face and the face is in dark shade or bright sun you'll probably get a good exposure of the face and the rest of the photo will "be where it will be", and that is okay. If the face/subject is in bright sun and the background is dark the histogram packs to the left side, period. If the face/subject is dark and there is a brilliant sunrise behind, the face/subject will still be fine but the histogram will pack to the right side.

If you matrix meter in camera (normal for landscapes) the histogram will center up and "look perfect" but your true subject, that face you were trying to get, may be too light or dark to use.

As a matter of course I normally shoot a little under to keep more detail in the clouds and wedding dresses.

Hope that helps.