If I have a picture I need to manipulate, say I need to resize it, does fiddling with the dpi settings do anything for me?
Say I start with an image that’s 1280x800 at 72 dpi, and I need to crop part of it out and blow it up. What should I do to maximize the quality of the new image?
Don’t do any resampling.
That’s the short version.
There are techniques that you can do when making large prints from low-res images to minimize obvious “jaggies”, but basically, all they are doing is blurring the photo. Once you have a digital image, you have all the information you will ever have - you don’t want to reduce it by resampling (making fewer pixels), and increasing the number of pixels doesn’t give you any more information than you already have. DPI is mostly meaningless, it’s the total number of pixels in each dimension that is important.
I realize this is a tangent, but since you mentioned blurring: I think I remember once reading a tutorial on using screen caps to make custom DVD cover art which suggested that after you’ve got all the images where you want them, to sharpen everything a few notches and then blur it all a few notches to minimize the obvious cuts between different images. Or possibly the other way around?
Is this a reasonable practice, considering that I’m not talking about commercial art, just stuff for my own use?
It’s a matter of personal taste.
I’ve been to Photoshop presentations where the instructor goes through dozens of steps - sharpening, masking, brightness / contrast adjustment, etc. to end up with the image that they like. When they compared the result side-by-side with the original, I often like the original better. You can’t make a silk purse out of a sow’s ear - if you are working with screen caps, I would play around with your images and see what looks best to you. My wife derides me as being too much of a purist - I particularly detest the current trend towards Kinkade-style prints, where all the colors are neon-bright. Still, that’s what seems to sell.
The DPI is just a field in the file which tells the program doing the printing what is the intended DPI for printing and the program can use it or ignore it but it does nothing to the picture itself on the computer where pixels are the only meaningful units.
So to go back to my custom DVD cover art example, if my printer is capable of 600 dpi, does upping that field from 72 to 600 actually do anything for me, or does it just cause the image to print at the wrong size on the page?
The only correct answer is “it depends.”
Most modern DTP programs (InDesign, Quark, etc.) don’t really care what the DPI field says. When you place the image, you scale it to the size you want on the print - the program does all the DPI calculations internally. If you change the DPI on a linked image, the program may print the image bigger or smaller, but you could just re-size it to compensate - the quality will remain exactly the same.
Some of you are using the term “dpi” (dots per inch) correctly; and others of you are really talking about “ppi” (pixels per inch). Not at all the same thing.
A digital picture usually represents an image with a rectngular array of picture elements or pixels of uniform size and spacing. One of the limits to picture quality is that details that would occupy less than one pixel can’t be represented usefully. If you crop your image you are keeping only the pixels within some region.
There are various means of resampling this array so that there are more pixels afterwards. It isn’t just blurring; there are several ways of doing this that can be better than blurring, at least in some respects. But none of them are way better.
The digital picture is an informational abstraction, but to show it on a screen or print it on a page the pixels have to be mapped onto a real physical background, so the pixels must take on a physical size. Specifying dots per inch or pixels per inch has the effect of specifying a physical size.
Panache, what do you mean? If the dots are picture elements, they are pixels. Are you talking about some system like halftone printing where there is a dot pattern that is distinct from the picture element dots?