Intro:
I've been doing HDRs for only a few months, but I've tried at least 30-40 images in that time.

Basically HDR stands for High Dynamic Resolution/Range. It's all about taking many of the SAME shots at different exposure and then overlaying the shots to create one. From the 'many' shots, only the properly exposed parts of the photo is used, and the rest is discarded. For this reason, a tripod is a must and overlay will not work unless pretty much every pixel lines up. Some ppl experiment with changing exposure via RAW exposure software, but it doesn't work (as well).

The human eye's dynamic resolution is massive and that's why we can see the bright sky AND dark shadows without one being overexposed or other being underexposed. Unfortunately camera sensors aren't as good, and the only way we can achieve this is to use HDR.

Number of shots and metering:
The minimum number of shots taken is two, but I usually do 3, 5, 7 or even 9 shots. Depending on the number of shots, I bracket the ev compensation with less ev for more shots.

Bracketing is done via camera controls and most canon/nikons have the feature albeit 3 shots for most entry level DSLRs. I'm not sure about Canon, but Nikon have a 'BKT' button which you depress and then spin the thumb dial to choose the number of shots. The forefinger dial is used to set the amount of ev to space it out across. If you don't have this feature, then you can do it manually and expose for different parts of the photo, (remember to not move that tripod).

If the scene is pretty simple and there's only 3 areas of varying exposure, then I will meter for the three distinct areas. For example and in many cases, the three areas will be:

1) The sky, which will require less exposure: -1.0 ev
2) The foreground, which will require a longer exposure to lighten it up more: +1 to +2 ev
3) The middle ground - or usually point of interest - which is usually the 'correct' exposure based on spot metering: 0 ev

So in general, a good starting point is three exposures about 1.0 ev apart. (ie: -1 : 0 : +1). Sometimes the foreground will need to overeposed +2 depending on lighting.

If there's more dynamic range, then I will go for more shots. I've used up to 9 in some instances, but I think that's overkill. In case of 7 or 9 photos, I use less ev, usually about 1/3 ev apart. Beware that the more shots you use, the more harder it is to overlay, and more blurring and noise will be created.

My camera, however, allows me to do 9 shot exposures and sometimes I take all 9 at 1.0 ev apart, and when I get back to the computer I select the three shots that best expose for areas I'm interested in.


Software:
I use "photomatrix pro". You can download a free trial product from their website here, but it contains a watermark until you purchase it.

I'm aware you can also use PS as well, but I don't know how to use it and when I first did my research, I was told in many instances that Photomatrix was a lot better apart from the extra cost.

Method:
Okay, I'll use the example I took last night. I've taken three of the following shots.

1) This image is at -1.0 ev. Underexposed to expose for the sky. The basic premise being less light is allowed to the sensor to allow for correct exposure of low level sky and sunset areas. It also correctly exposes for some lights on the buildings.




2) This is the normal exposure spot metered on the clouds, and hence the clouds are correctly exposed. The buildings are a tad bit brighter and the street is lit up.




3) The third image is shot at +1 ev to allow for exposure of the buildings. As you can see, it isn't overdone as is in many cases of HDR, and there are still some heavy shadows present. This is desirable as overdoing it (exposing the foreground perfectly) is unrealistic.




Now that I have the three images to work with, I open up the Photomatrix software and go to "HDR" and go down to "generate".



It will then ask to locate the files of which you choose all three (holding down the "control" button and clicking on them). Then click "OK".

It will lead to this window here:



Click the box to mark "allign sources..." and then click "OK".

Once you've done that, the software will churn thru the three photos and merge them. This process may take a while depending on the size and complexity. You will get something weird looking like this:



Don't worry about how it looks right now, as you still need to convert it to 8 bit. To do this, click on "HDR" and then go down to "tone mapping" as follows:



Once you have clicked on that, you will get to a window that looks like this:



In this window you can play with the different settings for strength and saturation. I find that whatever it leaves it at is best. If you want to overdo it and get the cartoon effect going, then this is the window to play with the settings. Make sure it's on 8 bit resolution and click on "OK".

You will get the end product which looks like this:



Save the image.

From here, I go back to Photoshop and edit it further to reduce the HDR effect. Usually this is done by darkening the picture slightly using levels and curves. I then go thru all the normal post processing tools to make the best image I can.

Finally, a run thru a noise filter is handy, as HDR creates a lot of noise when overlaying images.

The final image: