Ad
Ad
Ad
Pages: [1] 2 3 »   Bottom of Page
Print
Author Topic: resizing images  (Read 13604 times)
abaazov
Full Member
***
Offline Offline

Posts: 173



WWW
« on: February 05, 2006, 09:49:20 AM »
ReplyReply

from what i understand, there are many ways to resize an image. i was wondering what would be the best way to make an image smaller. for example, i have an image that is 12x18 with a ppi resolution of 240. i want to print it at 11x17. i can increase the ppi to around 265, and that will give me the print size i want. is that better, worse, or the same as resampling the image to fit into 11x17?
thanks....
amnon
Logged
michael
Administrator
Sr. Member
*****
Offline Offline

Posts: 4924



« Reply #1 on: February 05, 2006, 09:57:42 AM »
ReplyReply

You'll find as many answers to this as their are photographers.

My rule of thumb, is that when I resize for printing I let the PPI go as low as 240 without worrying about it.

Below 240 I will use Photoshop's Bicubic Smoother to set the output to 240 at whatever size I want.

It used to be that some people would upres in small increments, but this was rendered unnecessary when the Smoother function was introduced.

Some people believe (or have found though their own testing) that products designed for upressing do a better job that Bicubic Smoother. The last time I did my own tests (a couple of years ago) I didn't find a single one that did visibly a better job on prints at normal viewing distances).

Michael
Logged
abaazov
Full Member
***
Offline Offline

Posts: 173



WWW
« Reply #2 on: February 05, 2006, 10:00:27 AM »
ReplyReply

thanks michael

amnon
Logged
Tim Gray
Sr. Member
****
Offline Offline

Posts: 2002



WWW
« Reply #3 on: February 05, 2006, 10:33:57 AM »
ReplyReply

What's nice about Qimage (PC only) is that you don't have to worry about resizing at all.  It's uprezzing technology is at least as good as PS and it optimizes dpi on the fly.
« Last Edit: February 05, 2006, 10:34:29 AM by Tim Gray » Logged
jmccart
Newbie
*
Offline Offline

Posts: 24


« Reply #4 on: February 05, 2006, 12:50:35 PM »
ReplyReply

 
I have used Genuine Fractals for resizing  and love the results.  It provides the most direct approach for me.
Logged
Mark D Segal
Contributor
Sr. Member
*
Offline Offline

Posts: 7126


WWW
« Reply #5 on: February 05, 2006, 01:54:44 PM »
ReplyReply

Amnon,

I like Michael's advice on this. I do likewise.

That much said, there are a few additional comments to address the questions you asked.

Firstly, in the case you asked about, you are downsizing, not up-sizing. If you wish to have a specific image resolution (PPI) - necessary for some things, but not others - when you are downsizing, in the Photoshop Image Size dialog box you would select "Bi-cubic Sharper" rather than "Bicubic Smoother". The former is for downsizing and the latter for up-sizing.

Like Michael, normally when I change the linear dimensions of an image I leave "Resample Image" UNCHECKED, I change the linear dimensions to what I want and let the PPI resolution take care of itself, as long as the PPI doesn't sink much below 240 when I am up-sizing. This is because it is best for image quality to avoid resampling as much as possible and 240 produces a very acceptable large format print (at least on Epson 4000/4800 printers where I know from personal experience).

Now, since you are REDUCING the linear dimensions in the example you asked about, when you leave "Resample Image" unchecked and change one of the linear dimensions (with "constrain proportions" on), the PPI will increase, which is fine. If you reach a point where PPI increases beyond 480, you may wish to resample down to this maximum value, because sending more than 480 PPI to the printer really achieves nothing - you are hyper-rich in resolution.

In the example you gave, moving from 240 to 265 by just changing the linear dimensions with "Resample Image" UNchecked is theoretically the best of all worlds because you are getting more resolution without any quality degradation from the resampling process - though I agree with Michael that Photoshop's new resampling algorythms are so good that within a reasonable range you won't really notice the difference.
Logged

Mark D Segal (formerly MarkDS)
Author: "Scanning Workflows with SilverFast 8....." http://www.luminous-landscape.com/reviews/film/scanning_workflows_with_silverfast_8.shtml
Kenneth Sky
Sr. Member
****
Offline Offline

Posts: 422


WWW
« Reply #6 on: February 05, 2006, 10:26:31 PM »
ReplyReply

Scott Kelby adds an interesting wrinkle to the process of upwards resizing in The Photoshop CS2 Book (p.109). He suggests using Bicubic Sharper gives better results.
Logged
Mark D Segal
Contributor
Sr. Member
*
Offline Offline

Posts: 7126


WWW
« Reply #7 on: February 06, 2006, 08:29:22 AM »
ReplyReply

I've seen that too, but I haven't bothered to test it because I re-res so seldomly and get great results using Adobe's intended tools when I do; that said, Scott most likely has tested it, so he may well be right, but I wonder in what circumstances the difference would really become noticeable.
Logged

Mark D Segal (formerly MarkDS)
Author: "Scanning Workflows with SilverFast 8....." http://www.luminous-landscape.com/reviews/film/scanning_workflows_with_silverfast_8.shtml
p.tinson
Jr. Member
**
Offline Offline

Posts: 58


« Reply #8 on: February 06, 2006, 12:50:05 PM »
ReplyReply

Another method I have seen advocated is to upsize to about 5% larger using Bicubic smoother and then downsize to the required size using Bicubic sharper.
This was stated to be almost as good as using fractuals in Qimage.
I have used this method to enlarge a 6.1 mega pixel camera image to A1 size with very acceptable results.

Peter
Logged
jani
Sr. Member
****
Offline Offline

Posts: 1604



WWW
« Reply #9 on: February 14, 2006, 08:34:58 AM »
ReplyReply

I guess I'm the odd one out; I prefer bicubic smoother to resize to smaller images, but that's usually for web usage.

Maybe I should try using sharper for a while to see if I'm more pleased with the results now than I used to be.
Logged

Jan
abaazov
Full Member
***
Offline Offline

Posts: 173



WWW
« Reply #10 on: February 15, 2006, 12:43:46 PM »
ReplyReply

what about resizing in camera raw? is that a good/bad idea?
Logged
Brian Gilkes
Sr. Member
****
Offline Offline

Posts: 431


WWW
« Reply #11 on: February 15, 2006, 02:01:18 PM »
ReplyReply

In theory I agree with the " don't resample" advice. In practice when printing for clients, I ask for files to be sized at 360ppi. This avoids having to deal with resolutions way outside acceptable limits.
There is a school of thought that suggests that if the native resolution of the printer is , say 720 ppi, as I believe the Epson 800 series is, then resolution of file should be 180, 360 or 720. Maybe the printer doesn't have to think too much then! With 180 below general acceptability and no advantage of 720, then 360 remains. I have not noticed , nor has any comments, regarding loss of acuity by resampling to 360ppi. Knowing how to sharpen properly seems much more important and I always ask for files that have no sharpening at size applied so that I can do it. I think there could be a book written on sharpening. Further I sharpen on a seperate layer on a duplicate of the original and would have a duplicate for each size to be printed. Sharpening would most likely be different for different sizes.
I use the PS Bicubic options for resizing. For large shifts I still move 10% at a time, which may be just habit, but a number of respected printers still advocate this approach even with the recent algorithms.
As an aside , I am at the moment experimenting with printing at 720 ppi  a  vector file from Illustrator that I have converted in Photoshop. I realize rasterisation in a RIP is the preferred approach, but I am awaiting some results of RIP comparisons before lashing out on one for the 9800.
The vector file requires considerable upressing and there are small jaggies at 100%.I would welcome comments from anyone with experience in dealing with this sort of file.
Cheers
Brian
Pharos Editions
www.pharoseditions.com.au
Logged
Stephen Best
Guest
« Reply #12 on: February 15, 2006, 04:43:49 PM »
ReplyReply

Quote
There is a school of thought that suggests that if the native resolution of the printer is , say 720 ppi, as I believe the Epson 800 series is, then resolution of file should be 180, 360 or 720.

My understanding is that it's 360 for the 48/78/9800. Selecting "Finest Details" bumps it to 720. The smaller desktop models are all supposed to be 720.

I also go with 360 (exactly). If the file size is close to what's required, I up/downsample with vanilla Bicubic. If the change is somewhat larger, I use Bicubic Smoother (upsample) and Bicubic Sharper (downsample). In all cases, immediately prior to printing I apply fixed output sharpening optimized for 360 (using an Action based on TLR's scripts - http://www.thelightsrightstudio.com/). I think the differences with a range of resolutions are however likely to be subtle, but giving the driver exactly 360 is the optimum ... in theory. If your files sizes are always limited you may want to consider a similar workflow based around 240 (exactly).
Logged
Lisa Nikodym
Sr. Member
****
Offline Offline

Posts: 1702



WWW
« Reply #13 on: February 15, 2006, 05:12:37 PM »
ReplyReply

I did some actual experiments with my Epson awhile back, and have reiterated the results here a few times since then, but it sounds like it's time to repeat them for newcomers:

There is no discernable difference between 240 ppi and 360 ppi.  180 ppi is slightly worse (but you have to look very closely to see the difference).  Sending it to the printer at the native resolution without resizing (in my particular experiment, about 215 ppi) was noticeably worse than any of the preceding cases, even than the 180 ppi case.  This indicates that the printer's internal interpolation routine is worse than Photoshop's.

Calypso Imaging suggests that you can start with images down to about 150 ppi and still have them come out "good", though not excellent.  For excellent, you need more like 240 ppi.

Lisa
Logged

Mark D Segal
Contributor
Sr. Member
*
Offline Offline

Posts: 7126


WWW
« Reply #14 on: February 15, 2006, 06:17:48 PM »
ReplyReply

I very seldomly resample images. I change the linear dimensions and let the PPI change accordingly (meaning all the original pixels remain, only their size changes to accommodate more or fewer of them per inch of output), as long they remain within a range of 180 to 480. Quality is usually acceptable at 180 or even a bit below. Between 240 and 480 you have to look very hard and use some imagination to discern quality differentials on prints (in my case printed on Epson Enhanced Matte using an Epson 4800 with K3 inks). The integer divisor issue and the printer's so-called "native resolution" issue may be interesting theory, but they are heavily over-rated in terms of their operational signifigance.
Logged

Mark D Segal (formerly MarkDS)
Author: "Scanning Workflows with SilverFast 8....." http://www.luminous-landscape.com/reviews/film/scanning_workflows_with_silverfast_8.shtml
Lisa Nikodym
Sr. Member
****
Offline Offline

Posts: 1702



WWW
« Reply #15 on: February 15, 2006, 06:25:12 PM »
ReplyReply

Quote
The integer divisor issue and the printer's so-called "native resolution" issue may be interesting theory, but they are heavily over-rated in terms of their operational signifigance.

I did the experiment myself, and, I assure you, the integer divisor issue has a significantly greater effect than the 180 vs. 360 issue.

Lisa
Logged

drew
Sr. Member
****
Offline Offline

Posts: 477



WWW
« Reply #16 on: February 15, 2006, 06:57:51 PM »
ReplyReply

I am totally with Tim on this one. Resizing for printing is just a pfaff. This is why I used to sometimes use CS2's print with preview controls to fit the image to a particular print size. Lets say I had converted a 1DS MKII raw file with Capture One to an output resolution of 300. This would give print dimensions approximately conforming to about an A3+ sheet of paper. Therefore, CS2 must downsample this file to print onto A4 paper. Not a problem you would think for such an expensive application with all of its sophisticated tools. However, I kept getting aliasing on sharply defined edges and I could just could not understand why. Qimage on the other hand uses sophisticated algorithms to resample images for printing (?Lanczos interpolation). Absolutely perfect prints of any size every time and no need to alter the original converted Tiff. Qimage also does not run out of steam like Photoshop when printing reallly big images. It really is a nice program and excellent value for money. Next best thing to a RIP and at a fraction of the cost.
Logged

Andrew Richards My Webpage
Mark D Segal
Contributor
Sr. Member
*
Offline Offline

Posts: 7126


WWW
« Reply #17 on: February 15, 2006, 07:54:01 PM »
ReplyReply

With my Canon 1Ds as well, something has to give squeezing all that information into an A4 or letter-size image. I simply don't resample. If I use the full frame and dimension the printed image area to Letter-size paper (just a bit different from the UK's A4) I do a 9*6 inch image and that yields 451.556 PPI. I've tested alternative outputs by cropping the image to provide 360PPI or I've re-rezzed the whole thing to 360 PPI using bicubic sharper. I couldn't see the difference on the prints.

Likewise, an A3 from full frame is 14.5 by 9.65 inches of printed area at 280.276 PPI, and I find it very difficult to detect quality difference whether one prints this at 240, 280 or up-rezzed to 360.

I have had quality problems up-rezzing severe crops to fill an A3 from something as low as 120 PPI starting to 240 target - here one sees some brittleness in the resulting image texture. I'm yet to try QImage, though I keep hearing great things about it. Perhaps testing results depend partly on the nature of the image being tested, the printing environment, how sharp one's eyes are, etc. Likely quite a few variables.

Anyhow, for what it's worth I collaborated with Harald Johnson on resolution settings for scanning film (where the integer divisor debate is also alive and well) and we came to similar conclusions (there is alot of latitude) - this is reported in his Mastering Digital Printing Second Edition pages 83 to 85).
Logged

Mark D Segal (formerly MarkDS)
Author: "Scanning Workflows with SilverFast 8....." http://www.luminous-landscape.com/reviews/film/scanning_workflows_with_silverfast_8.shtml
Stephen Best
Guest
« Reply #18 on: February 16, 2006, 03:58:40 AM »
ReplyReply

Quote
There is no discernable difference between 240 ppi and 360 ppi.

Try this. Create a new image at 240ppi, white canvas, draw a 300x300 black pixel square and add some text in the middle, say Helvetica 3pt. Create a second image, this time at 360ppi, draw a 450x450 pixel square (same corresponding size) and add the same text, again at 3pt. Print both out on Enhanced Matte (or similar) at 1440dpi unidirectional. Compare. My eyes aren't great these days but I can see a difference both in the weight of box and the resolution of the text. Without a loupe. The same would apply for any micro details in your files ... assuming the information is there in the first place.
Logged
drew
Sr. Member
****
Offline Offline

Posts: 477



WWW
« Reply #19 on: February 16, 2006, 05:55:26 AM »
ReplyReply

Stephen,
I am sure you will see a discernable difference. What I also urge you to try is leaving your test file at 360ppi and sizing the print job using print with preview in Photoshop and then doing the same with Qimage. I am fairly certain you will be impressed by the difference.
Logged

Andrew Richards My Webpage
Pages: [1] 2 3 »   Top of Page
Print
Jump to:  

Ad
Ad
Ad