*** For details on how to be removed from this list visit the *** *** CCP4 home page http://www.ccp4.ac.uk ***
On Wed, Mar 29, 2006 at 07:58:56PM -0800, Edward A. Berry wrote: > > Tim Fenn wrote: > >> > >>Say you want a fig. with final resolution of ~ 300 dpi at a final size > >>of 4-in by 3-in. > >> > > > > > >I always notice people do this, but it doesn't actually get around the > >issue of print resolution versus image resolution (and no, the two > >aren't the same). Just becuase I have a 1200x900 image scaled in a > >4x3" box doesn't mean my printer will use 300DPI when it prints - I've > >seen cases where printers/programs actually take a 72DPI image and > >upsample it to 300, then rescale it to the desired size on the page, > >which doesn't look too good in print. Anyway, I've started using the > >GIMP to reset my print size without affecting the image in any way. > > If N. Stanley takes his images which are unacceptably low > in resolution, but about the right size, and uses gimp to > change them to higher resolution, he is either going to get > very small images (if gimp does not "resample" but simply > redefines the size of the pixel), or he is going to get a > picture which nominally has higher resolution and takes up > n**2 times as much disk space, but looks identical to the > original (if gimp resamples). If he wants a 4x3 picture > that really has 300 dpi resolution, he has to start with > 1200x900 pixels, no matter how painful the math. > All it does is change the image header, and leaves the data completely untouched. The dpi value in an image header can be whatever the heck you want - for the most part, its irrelevant and ignored (the only reason alot of programs stick 72dpi in there is because thats the average video screen resolution). Just because I make a 1200x900 pixel image says *nothing* about the dpi. > For me the primary characteristic of an image is the > number of pixels. This determines the information content, > the level of detail. You can print it at 2 ppi on a billboard > and look at it from across town, or print it on a postage > stamp and look at it with a magnifying glass, the same level of > detail is present as long as you keep the same pixels. Sure - but my point was that some programs/printers DO read in the dpi value in the header, and end up doing strange things with it. For example, if I make a 1200x900 pixel image in povray, it arbitrarily sticks in 72 dpi in the image header. So if I dump that image into microsoft word and hit print, will it use the value in the image header and do some kind of weird resampling, or be smart and figure it out from the size of the image in the document divided by the pixel size of the image? What if that number isn't useful for the printer, how is it resampled? If I resize the image, what then? I guess I could read the source code - oh wait... > Some image formats (like GIF) don't even record the > image size/resolution, but leave that up to the displaying > program/device. > If you're referring to image resolution, then thats just taken as the size of the image... and it *IS* up to the displaying program/device how to render something: image resolution != display resolution, and thats where things get confusing. My monitor has about 110 dpi width @1600, so a 1100 pixel wide image covers 10 inches, but may print as a completely different size, depending on the software and the printer. Thus the reason the "dpi" information in the header can be important, since some software will use it at face value. > The go-between is the editor (or reviewer or whoever decides > what size the picture should be). If he says "No you can't > have a full page for that figure, you have to make it > one column width" then you look up the column width and > multiply by the resolution the publisher wants to get the > number of pixels you need. Make a figure with that many > pixels, the resolution is irrelevant. This gets complicated, however - not all journals provide all the information you need to figure this sort of thing out, and some just go ahead and resize images without even paying attention to your image size and their print resolution, much less the dpi stored in the image header. > Then just before you you upload it to the publisher, convert to a > format like tiff that does preserve figure dimensions, and set the > resolution without resampling the pixels. And voila- the size will > be one column width, and all the original pixels created by the > graphics program are there. > Ick - TIFF is an evil, encumbered format. Use PNGs (which also use more efficient and free lossless compression algorithms): http://en.wikipedia.org/wiki/Png Which reminds me: I've even seen journals ask for image formats that are lossy by design, like jpgs!!! -Tim -- --------------------------------------------------------- Tim Fenn [EMAIL PROTECTED] Stanford University, School of Medicine James H. Clark Center 318 Campus Drive, Room E300 Stanford, CA 94305-5432 Phone: (650) 736-1714 FAX: (650) 736-1961 ---------------------------------------------------------
