Question about file sizes.

[Cassandra]

If I'm shooting jpegs in large, superfine mode, why does the MB size vary wildly? A wide scene might be 3+ MB or might be smaller. A macro scene might be 1 or 2 MB or it might be larger. Do I have any control over the file size coming out of the camera (P&S)?

[JeffR]

1. There's a relationship between the amount of detail & file size

Specifically, I think it has to do with both the number of different colors the sensor uses to capture the image, and the amount of color differentiation pixel-to-pixel. Luminosity difference pixel-to-pixel is also a factor, I believe. The only control you have, as I understand it, is setting the file size mode in the camera; whatever mode you're in, you'll still have file size variations.

[TahitiNut]

2. Yup. It's due to the compression algorithm and the amount of "information" needed.

You can actually experiment. If the photo were a completely uniform color and luminosity (a bedsheet), the resulting file size at a specific compression level would be very small, while a photo of a floral bouquet, for example, would be quite a bit larger.

As a mental exercise, imagine the number of words it'd require to describe the photo to a skilled artist in such a way as that artist could come up with a replica. In a nutshell, that's the underlying information theory driving folks in the efficient digitalization of a photo. There's the brute force of pixel-by-pixel and then there's the intellectual comprehension of the amount of detail (=information).

We should ALWAYS remember that DATA does NOT equal INFORMATION any more than TALK equals INFORMATION. (It shouldn't take more than 5-10 minutes watching TV to know this.) The challenge, as always, lies in optimization.

:silly:

[Cassandra]

3. This has also come up on a project I help out with...

taping the leader of my women's group. Each 1 and 1/2 hour tape (we're upgrading to a camcorder with a 60 GB HD, YAY!) becomes after download about 250MB, so that many will fit on a 4+GB DVD which says it's only 2 hours. The trick seems to be that the camera doesn't move and the leader moves very little.

[TahitiNut]

5. Yes, the current compression algorithms don't handle pan or zoom very well.

You'll get far less efficient compression when the camera pans ... when the frame changes in relation to the background. While we can sense intuitively that the "information" content of the frame isn't changing that much, the algorithms have heavy reliance on the frame geometry. Some experimental approaches attempt a more frame-free context and we'll eventually see more efficiency in that regard.

So, for administrative/logisitcal/storage purposes (i.e. smaller file sizes), it's far better to have a pair of cameras with a fixed/stationary frame, moving/panning/zooming only one at a time with planning for direction of movement so the video can be A/B edited and the result be a series of stationary frames in which the action moves. (Think in terms of overlapping windows in a wall through which some action is played out.)

It's difficult for me to explain the technique without sketches and/or acting out. But it's how technology drives technique ... a factor in many skill-based activities.

[Cassandra]

4. Thanks. Informative yet frustrating.

I was poking around wondering if submitting to stock photography sites might be worthwhile (ignoring the fact that I'm thinking of this with a point and shoot camera) and they seem to want minimum 4MB photos . I'm trying to pull money together to get a better camera (I'm thinking Canon EOS Rebel XSi; it's not so heavy that it will be left on the table when I go out) but, as usual, you have to have some money to make some money.