| Author | Thread |
|
|
12/05/2003 02:50:04 AM · #1 |
Hi, I have noticed that most people with 5 megapixel cameras are offering prints up to 16*20 inch sizes. While at 150 dpi, the maximum size print of a 2560*1920 file can be up to 17*12.8. Does that mean everyone is resampling up their images? Is there not a loss of quality in the prints? I would be grateful if someone could explain this to me. Thanks.
Message edited by author 2003-12-05 07:50:50. |
|
|
|
12/05/2003 03:04:46 AM · #2 |
My Sony is a 3.3 megapixel and I was able to get 16x20 prints, or 20x30 respectively, out of it. It was accomplished through resampling to 150dpi. Even on my 6.3 megapixel I have to resample up to 150dpi for 16x20 and 20x30, although not as much. I have ordered a 10x15 from the Sony and it looks great! The admins won't approve an image that has been resampled too much and loses quality. I hope this helps.
Edit: As Eddy suggests - I use the 'stair interpolation' method myself - works great.
Message edited by author 2003-12-05 08:06:45. |
|
|
|
12/05/2003 03:04:54 AM · #3 |
You must have a minimum of 150DPI at your desired print size to sell the image publicly. If you just want to purchase a 20x30 print that is 100dpi for yourself, you can do that without any restriction. In all cases, DPCPrints handles the necessary upsampling to "printer" resolution automatically using algorithms optimized for the output device.
So if you want to publicy sell a 16x20 print but your image is only 118 dpi, you need to upsample. John covers "stair interpolation" (which usually yields decent results if you don't need to upsample too much) in his tutorial.
There are also third party programs like LizardTech's Genuine Fractals and Extensis' pxl SmartScale that are designed to do upsampling. I personally get satisfactory upsampling to 150 dpi with stair interpolation, so I just use that -- it's free.
You may also find this article about creating bigger prints informative.
Message edited by author 2003-12-05 08:13:10. |
|
|
|
12/05/2003 05:37:20 AM · #4 |
Originally posted by EddyG:
I personally get satisfactory upsampling to 150 dpi with stair interpolation, so I just use that -- it's free. |
That's free if you already own the expensive photo-editing software program. |
|
|
|
12/05/2003 06:29:21 AM · #5 |
The newer method being touted by gurus (apart from buying other software -- some really like SmartScale) is simply to use Photoshop CS's new Bicubic Smoother when sampling up. (Bicubic sharpen when sizing down.) I have yet to do comparisons, but apparently using bicubic smoother in a single step is better than using the bicubic stair method.
I only throw that out as a test it yourself and see... |
|
|
|
12/05/2003 06:37:20 AM · #6 |
Originally posted by coolhar:
Originally posted by EddyG:
I personally get satisfactory upsampling to 150 dpi with stair interpolation, so I just use that -- it's free. |
That's free if you already own the expensive photo-editing software program. |
The TECHNIQUE is free and functions with any (free) editing program which can resample at all. |
|
|
|
12/05/2003 06:38:28 AM · #7 |
Originally posted by Patella:
The newer method being touted by gurus (apart from buying other software -- some really like SmartScale) is simply to use Photoshop CS's new Bicubic Smoother when sampling up. (Bicubic sharpen when sizing down.) I have yet to do comparisons, but apparently using bicubic smoother in a single step is better than using the bicubic stair method.
I only throw that out as a test it yourself and see... |
I just got this at work. What's a "fair" test -- a landscape upsampled to 200% of original? |
|
|
|
12/05/2003 07:05:41 AM · #8 |
| Thanks for clarifying that General. The tutorial linked to in the same post that said it was free spoke of using a $15 plug-in to PS 5 or higher. Apparently the tutorial is out-of-date. |
|
|
|
12/05/2003 07:13:35 AM · #9 |
| The tutorial references one of Fred Miranda's Photoshop actions, which he does indeed sell on his site. His actions automate many common digital darkroom processes, and are optimized by Fred Miranda for "best results". Many people pay for and use his actions, and speak quite highly of them. |
|
|
|
12/05/2003 07:26:55 AM · #10 |
Originally posted by coolhar:
Originally posted by EddyG:
I personally get satisfactory upsampling to 150 dpi with stair interpolation, so I just use that -- it's free. |
That's free if you already own the expensive photo-editing software program. |
Its free in plenty of free tools too. stair interpolation is just a process, not a product. Bicubic resampling is available in anything that could be considered a serious photo editing tool (see my profile for about 15 free or budget versions, for a variety of OS platforms)
Basic principle is that the interpolation errors of doing a 100% increase in one step is larger than the cumlative interpolation errors for doing 10 10% increases.
While mathematically this may or may not make sense, the end result is that apparent detail is preserved (questionable if any more actual detail is preserved) so the result looks better.
As I learned a long time ago, when it comes to writing image processing alogrithms, what looks right, is right - no matter if the mathematics backs it up, or not. |
|
|
|
12/05/2003 08:16:58 AM · #11 |
| I don't go below 200dpi for lamda style printers, and never below 250 (i keep it at 300) for inkjets. I have interpolated my files upto 40"x50"'s. With fantastic results. I don'y shoot medium format anymore... |
|
|
|
12/05/2003 08:46:27 AM · #12 |
I'm such a novice to this topic and to putting up prints, I have to ask this: when I resampled the original (72 dpi) to 300 dpi, I did this in a single step as opposed to the suggested increments. The results, to my eyes, were good.
I am aware of the reasons for not doing it the way I did it, but what if the respective results of the two different processes are at no discernible variance, practically - or are they?
PS: I'm using Graphic Converter (Mac), not Photoshop, if software has any bearing here.
Message edited by author 2003-12-05 13:53:09. |
|
|
|
12/05/2003 09:04:19 AM · #13 |
you can only really tell what the results are like looking at the image at 100% magnification (anything else is being further resampled)
But, if at 100% magnification, you can't tell any difference, then there is no difference. How it looks is the final and only really valid quality test for digital images. |
|
|
|
12/05/2003 09:27:04 AM · #14 |
Hm, truth is, I'm not at all happy with the image at 100% magnification (either original or the resampled version). The largest print size (8x10), however, I want to make available would, roughly, correspond to a 33% magnification. At 33%, the image looks fine.
Do you feel, I should or could go with this? |
|
|
|
12/05/2003 09:31:34 AM · #15 |
Then I'm confused. Why resample it to 3 times the resolution you require?
8x10 at 300dpi is 2400x3000 pixels - are you saying you resampled to 3 times that, so you'd be printing at 33% of the final version ? Or are you saying that the 72 dpi screen representation is about one third the dot pitch of your print target, so you display it at 33% it looks about the right size ?
A 33% preview (or anthing other than 100% preview) has been resampled further (typically using quick/ dirty algorithms) and thus smoothes out a lot of fine detail that is visible in the final print.
Message edited by author 2003-12-05 14:32:47. |
|
|
|
12/05/2003 09:37:34 AM · #16 |
Originally posted by Gordon:
Then I'm confused. Why resample it to 3 times the resolution you require?
8x10 at 300dpi is 2400x3000 pixels - are you saying you resampled to 3 times that, so you'd be printing at 33% of the final version ? Or are you saying that the 72 dpi screen representation is about one third the dot pitch of your print target, so you display it at 33% it looks about the right size ?
A 33% preview (or anthing other than 100% preview) has been resampled further (typically using quick/ dirty algorithms) and thus smoothes out a lot of fine detail that is visible in the final print. |
Oops, I meant preview magnification (screen size of image trimmed to approx. meet dimensions of print size). |
|
|
|
12/05/2003 09:43:45 AM · #17 |
Originally posted by zeuszen:
Oops, I meant preview magnification (screen size of image trimmed to approx. meet dimensions of print size). |
The only problem then is that the preview magnification is itself, another round of resampling (to show it on the screen at that preview)
100% is the only 'pixel to pixel' view, everything else is resampled to show the preview, so is less reliable. Now, if you are printing it smallish (8x10 or less) you quite possibly won't be able to see this, but certainly a 100% magnification preview is the only reliable way to evaluate potential print quality (and to do sharpening too, for similar reasons) |
|
|
|
12/05/2003 09:50:44 AM · #18 |
Gordon, I'd extend your excellent advice to say "at least 100%". I often find that I need to preview at 200% to really judge whether there is a difference between two versions of a file.
LCDs with digital inputs are a different animal, but for CRTs, the monitor will usually add some softening effect at a 100% view unless it is operating well below its max resolution, so I find it better to increase the magnification and thus reduce the effect of the monitor's softness.
Message edited by author 2003-12-05 14:50:56.
|
|
|
|
12/05/2003 09:56:25 AM · #19 |
Originally posted by Gordon:
...100% is the only 'pixel to pixel' view, everything else is resampled to show the preview, so is less reliable. Now, if you are printing it smallish (8x10 or less) you quite possibly won't be able to see this... |
I'm in a russian roulette state of mind after reading this. :-)
... |
|
|
|
12/05/2003 10:02:48 AM · #20 |
Are there applications that can resample at huge percentages, say 600%, with high quality? How are giant posters of photos made that were small snapshots? Are they just bad quality?
|
|
|
|
12/05/2003 10:06:43 AM · #21 |
Originally posted by dsidwell:
Are there applications that can resample at huge percentages, say 600%, with high quality? How are giant posters of photos made that were small snapshots? Are they just bad quality? |
Have you ever been really, really close to a billboard ? Dots the size of golf-balls.
Apparent resolution is an issue of viewing distance as well as image quality. So most posters aren't really designed to be looked at from really close up (also you start to find that USM/ sharpening is very releated to print size _and_ viewing distance too)
Mostly though, huge, high quality reprints are often made from medium and large format, because the initial negative are so much larger that the realtive resampling is a lot smaller. |
|
|
|
12/05/2003 10:09:04 AM · #22 |
i think the confusion may be with the fact that the pixel dimensions change as you change the DPI, but the print size does not.
to print an 8x10 you would resize the image to print at 8x10inches - then change the DPI to say 288 is a number i tend to use. the pxel dimensions will be 4 times the original - but the print will be the same size with both.
to see realistic what the print will look like - the image editor would need to have the image at 100% size ( much larger apperance than 8x10 inches on the PC screen, because the monitor is only capable of a 72dpi output ) if you view the high res image at only 50% its possibly going to look cleaner than the actual print will turn out...
hmm - not sure that will help, but maybe ;}
|
|
|
|
12/05/2003 10:09:18 AM · #23 |
Originally posted by kirbic:
Gordon, I'd extend your excellent advice to say "at least 100%". I often find that I need to preview at 200% to really judge whether there is a difference between two versions of a file.
LCDs with digital inputs are a different animal, but for CRTs, the monitor will usually add some softening effect at a 100% view unless it is operating well below its max resolution, so I find it better to increase the magnification and thus reduce the effect of the monitor's softness. |
Yup, I'd concur - though, typically in those cases, no resampling as such is going on, the pixels are just enlarged - but anything below 100% involves down sampling on the image, losing info. |
|
|
|
12/05/2003 10:11:16 AM · #24 |
Originally posted by zeuszen:
Originally posted by Gordon:
...100% is the only 'pixel to pixel' view, everything else is resampled to show the preview, so is less reliable. Now, if you are printing it smallish (8x10 or less) you quite possibly won't be able to see this... |
I'm in a russian roulette state of mind after reading this. :-)
... |
Well, it isn't a hard and fast rule to define what is 'acceptable'. what I used to find acceptable in a print, is no longer acceptable to me - so it is difficult to give a conclusive answer.
However, the 100% preview is going to look much more like the final print than the 33% view, in all cases. |
|
|
|
12/05/2003 10:14:37 AM · #25 |
on this subject i tried to save a printable 20x30 @ 288dpi image last night
straight off the camera - minimal editing to it.
win98se - 512M ram 78% free - p4 1.8ghz
error - could not save as jpg. not enough memory
i tried numerous saving methods to no luck and ended up scaling the image down to 16x24. anyone else run into to this
is there a way w/o losing quality to save large images
a side note - i noticed that the 16x24 print was accecpted for a 20x30 print through dpcprints - but i have no way to tell how it will look... |
|
Home -
Challenges -
Community -
League -
Photos -
Cameras -
Lenses -
Learn -
Help -
Terms of Use -
Privacy -
Top ^
DPChallenge, and website content and design, Copyright © 2001-2025 Challenging Technologies, LLC.
All digital photo copyrights belong to the photographers and may not be used without permission.
Current Server Time: 04/09/2025 07:11:20 AM EDT.
