Compression vs image size?

[David H]

Excuse my ignorance - I'm new to this game!

Just got a Canon Ixus 850is and trying to understand the manual! Not quite clear of the sigificance of using the larger or smaller image options. Isn't taking a larger image at higher compression effectively the same as a smaller image at less compression?

When would one use one rather than the other? If I've got plenty of space on the card, is there any reason why I wouldn't always use max image size and least compression (c. 3Mb per pic I think)?

I realise that I'm getting more detail than I need for small prints or simply viewing on a PC screen, but if I've got the space, it gives me the option to zoom in on part of the pic if I want doesn't it?

In what circumstances would I want to use either a smaller image size or greater compression?

Any explanations gratefully received!

Thanks.

[Jon]

You're perfectly right that shooting at the highest possible resolution regardless of the size print you expect to make is the best plan. After all, if you were shooting at the smallest possible size and got the shot of the millennium of Elvis getting out of a UFO it'd be kind of hard to sell to any of the better papers, while at the largest size, you'll be able to zoom in and see the mind-control implants they installed.

Changing the image size does just that - changes how many pixels are used to represent the picture. So you'll be trying to squeeze the data from, say, a 5x5 square (original) into a 3x3 square; each pixel in the final, downsized image will represent about 2.8 pixels of raw sensor data.

Changing the compression works somewhat differently. JPEG compression saves the colour for pixels as a "run", say "5 pixels of blue #0000F8, 4 pixels of blue 30000F6, 2 pixels of blue #0000FA", where #xxxyyzz is the hex red, green, and blue values for the colour. At the absolute least JPEG compression, every value will be saved. As the compression increases, colours that are close together will be averaged out, so our 5x#0000F8, 4x0000F6, 2x#0000FA" may get stored as "11x0000F8", so you've lost some onf the nuances. This won't make much difference in a subject with lots of detail, but in a broad expanse of almost the same colour, you'll see "banding" where the transitions between different colours occur, since those transitions become more pronounced.

So changing resolution costs you the ability to see fine details; changing compression costs you smooth colour changes. If there are lots of fussy details, you're better off using more compression and more resolution (which won't compress as well); when you're looking at large expanses of about the same colour, you can get away with a lower resolution (since there isn't quite as much detail), and use minimum compression to opreserve the nuances. Of course, neither of these will give the maximum compression for their respective situaions; if all you're interested in is a minimal representation with poor quality but maximum compression, choose the reverse combination for each of those situations, but expect lousy pictures.

[David H]

Thanks Jon. That really does help a lot. I'll be ready for Elvis now - but didn't he already have mind-control implants?!

[Jon]

Thanks Jon. That really does help a lot. I'll be ready for Elvis now - but didn't he already have mind-control implants?!
Well if you couldn't see them they might claim you just saw one of the numerous Elvis impersonators, but with the implants . . .

:{)#