View Full Version : Digital photography
How many megapixels would you need to blow up a picture to 16x20 without loosing quality of the picture?
I have a 7.1 megapixel camera and blew up a pictures to 16x20 and it was a little distorted.
Read this thread for some possible clues
It appears that the OP, John Werczynsi, thinks that Large Format begins at the print stage and magically works backwards. Hey, get a clue; large prints begin with the original medium. Now what are you doing here, Mr. Werczynski?
I agree with jj. He should go to the photo net digital processing forum.
With the exception of one, with garbage replies like that, I can see why new people are turned off by the large format community.
John let me apologize for the junk above and try to answer your question. If you're shooting with a 7.1MP camera, I know you are using a higher end point and shoot. The sensors in these are small compared to standard DX sensor DSLRs from the likes of Canon, Fuji, Nikon, etc. While the sensor may have more pixels, because it is smaller than the normal DX or full frame sizes, you're getting far more noise in a print the size of 16x20. This is but one issue. Others include the quality of the lens the camera uses, iso used, etc.
In short, an 8MP sensor properly processed in a good converter (like Capture one or Rawshooter) can deliver a decent 16x24. But I would say that it requires a decent DSLR like the D2X from Nikon, the 5D or 1Ds MK2 from Canon to give a truly detailed, sharp, low noise print.
What Dave said.
Also, the upsizing method and it's place in the post-processing flow can influence the final quality as well. Not much, but every little bit counts at that level.
To the point: If you know you are going to upsize (upres) the image, this should be the first step in the flow. It should also be done incrementally, instead of in one big pass. Keep increasing the size by no more than 10% in each step until you get to the desired size. Then apply all the other steps and THEN apply sharpening (USM) as the very last step.
As a final note, you will also achieve much better quality if you shoot raw and do all your processing until the sharpening step on the 16-bit file (per channel) instead of an 8-bit.
And don't mind some of the grouches you will inevitably encounter here - we're mostly a bunch of grumpy, aging men (save for a few ladies) here and that kind of attitude is part of the local folklore. There are some good souls under the crust though, you just need to be patient.
If you're using an older version of Photoshop, then stairstepping is good. If you have the newer versions, I suggest the following routine:
- Presharpen the converted image ( Try USM 300, .3, 2 ), fade luminosity
- Interpolate using bicubic SMOOTHER, in one step.....but interpolate 20% larger than your desired print size (ie; if you want 16x24, interpolate to 28.8")
- USM to taste ( Try 50, 1.5, 2 to start), fade luminosity
- Downsample back to 16x24 using bicubic SHARPER
I have found that when converting my RAW files in Capture One, and using this interpolation routine (which by the way is also mentioned here:
....I get finer detail, much less haloing, better accutance.....but at the expense of a slight bit more noise. This noise though is typically visible only on screen at 100%, but not so much on print, and is far less than anything you would get with film.
This is a simpler version of the routine I use, which involves interpolating and sharpening LAB mode channels differently, and seperately.....but it still does a fine job. Play with the presharpening and post sharpening a bit as every camera as well as every image requires differing amounts.
While I too think you might find better information elsewhere in the digital and publishing
boards, you can simply try it out to see what happens. For the price a large format photographer
may pay for a sheet of film and custom processing, you could simply upload a file to
a pro lab and then see it in all its glory, then be the judge. You do deserve better than
a crusty answer though, as you may end up turning to ( or returning to! ) LF photography
after exploring the modern day legends and mystique of the megapixel!
IF you are willing to accept 300 pixels per inch as the original standard for quality,
and if your image needs no cropping, and if all the pixels in your megapixels are
"effective" you would need 4,800 by 6000 pixels in the most simple sense, or 28,800,000
pixels. If you use 1,000,000 for mega, then you would need 28.8 megapixels. If you
use 1,048,576 for mega, then you might say that you need 27.5 megapixels. Generally,
a 16x20 will look very nice at 240 pixels per inch output resolution, at normal distances,
or 18,432,000 working pixels. I seem to remember Canon bragging about finally being able to
make a decent 11x17 ( full page ) print with their flagship DSLR somewhere, so that might
be an indication of what one camera maker deems "without loss of quality".
Naturally, most digital cameras will not do 16x20 without cropping, and most digital cameras have pixels that are not "effective", so in reality, you might need a 22 megapixel camera to make
a very high quality "honest" 16x20.
"Loss of quality" may come from other factors such as the optics used to make the
image in the first place. Most small digitial cameras ( except for MF ones that happen to
be fairly small now, such as the H1 ), suffer from inferior optics. No amount of megapixels
will make up for mirror vibration, camera shake or inferior optics.
Quality, beauty and satisfaction are in the eye of the beholder or creator.
Dave, would you please devote a paragraph or two describing your upsizing techniques? Also, is it your invention, or is based on something I can read on the web?
Quality, beauty and satisfaction are in the eye of the beholder or creator.
But in this forum it all starts from LF film. Why are you all entertaining a poor soul who wandered in here with a purely digital issue? Look, he ain't starving or homeless! Send him away.
Gently, if you like.
Dave, thanks for the tip, I made a note of it and I'll try it as soon as I get another digital camera. I use the latest Photoshop, but old habits are kinda hard to "unlearn". I do presharpening as well, but was hesitant to recommend it because it compensates for the Bayer sensor and to be honest, I'm not sure if the digicams are using it or not. Besides, lots of first time users tend to overdo it (I did) and it would serve no purpose in this particular case.
But we better be careful discussing digital, lest we get our hides drummed out of this board :)
Do you suppose the title of this thread might help the people who would rather not read about things digital---will it help them steer clear? Anyway, this thread contains some good information for anybody, including large-format users, who wants large prints from digital files.
I'm just heading out for ribs....so the paragraph will have be done later. I have been doing that routine and ones like it for quite a while. I then happened upon the article quoted above & I think it explains things better than I can.
I'll go into more depth later....now on to the ribs......
The issue regarding megapixels and enalrgement can be misleading.
What is often much more important is the *actual pixel size* of the sensor/chip in the camera used. First off, is it a CCD or a CMOS chip/ At 7 megapixel, likely a CCD chip, but not all CCD chips are created equal.
then you have the quality of the chip itself - how prone it is to noise, hot pixels, etc.
these just are some of the reasons why two cameras, say both 5 meg or 7 meg, can have very different prices.
another issue is lens quality. I have what woudl be considered a low quality digital camera - "only 3 megapixel" as I ahve been told, but the lens is a 12 element lens with ED glass. I find because of the high quality of the lens, and becasue of the high quality of the CCD chip, it outperforms other 4, 5 and even 6 megapixel cameras.
The big problem is finding out. You may find, as i have, that once you start getting into some of the technical questions - "what's the pixel size", a lot of these bright eyed sales clerks start to falter.
One last thoguht - megapixels do not account for other factors such as tonality. A photograph can be very grainy but still have wonderful impact. One of my personal favourites that I ever printed is a 16x20 enlargement fomr a 35mm neagative- HSI at that. grainy as hell, but the ontality 9at least for me) is just great, and the overall effect of the coarse grain witht he "dreamy" look of th einfra-red film just works great for me in that specific shot.
so don't get too caught up in grain - film or digital - look for the tool that will give you the art that you are seeing in your head and want to put to paper.
Sorry to not write back sooner....but you can't keep me away from a nice rib dinner!
The process I mentioned above can can work just fine in RGB mode with all the color channels. I have taken it one step further. I convert to lab mode, and have the luminosity and the Alpha A & Alpha B channels worked on seperately.
The luminosity channel is processed as I mentioned above.....with the exception of the "fade - Luminosity" part of course.
The Alpha A & B channels are processed basically as follows:
- Presharpen 300, 1.4, 4
- Bicubic stair step to desired print size
The 3 different channels are then opened all at once in photoshop, and through layers are combined in to one image and converted to lab, and then RBG.
Working with Alpha channels is tricky as the results you see on screen rarely correspond to what you get in the recombined RBG image or on print. You get a feel for it after a while. The above settings are just a starting point.
Watch the following:
Sometimes presharpening is not best. Then simply leave it until after interpolation and then be a bit more aggressive. If presharpening, don't go higher than about a 1.5 pixel width as you'll get color bleeding into the luminosity channel after recombining. Keep you threshold above 3 at least....pref 4 or 5 to avoid increasing noise on the Alpha channels.
As I said, this does give you a bit more noise that can sometimes be visible on screen at 100%.....but this only rarely shows up on print. If it is a concern, I run the Alpha channels through Neat Image with a very light noise dusting. This makes it even more convaluted, but it can be effective & I see no reduction in detail.
In the end, what I see as a difference is that I tend to get what appears to be a more natural look to the image.....more film like if you prefer, as opposed to a video like look when people follow a very basic workflow. I find it reduces halos, helps with specular highlights, and in the end, gives me better output for the sizes I normally print from a 1Ds, which is 16x24 & 20x30.
Also, while this works well on Canon CMOS based sensors, I find that it does not work as well with Nikon CCD sensors. Nikon's images have a bit more noise to them.....which on its own is not a problem. When luminosity channels & Alpha channels are processed like I said on a Nikon image, I find about an extra 1 to 1.5 stop disadvantage in noise when compared to Canon cameras. Because of this, when working on Nikon images, I skip the seperate Alpha processing and just work with the original flow I mentioned a few posts up.
Sorry to be wordy, but there ya go.
Dave, thanks so much! I wonder why you sharpen the A and B channels (with CMOS images) at all. I also wonder at the stair interpolation of A and B. I thought sharpening these two channels was seldom needed or wise. What goes wrong if you simplify these away?
I sharpen the Alpha channels with a larger pixel width to avoid noise. Using the values that I do on the lumi channel would increase color speckling on the alphas. By selecting a lower % but higher pixel width, I find a more natural appearance in the final product.
As to stair stepping, I found that because their is less detail info in the alpha channel, I am able to be a bit more aggressive on the interpolation part to maintain what detail there is, without having to worry about halos, jagged edges, or noise. While stair stepping can show more noise and haloing on the lumi channels, it doesn't appear to have that effect on the alphas. This final step with working on the alpha channels is subtle at best, and invisible most of the time. But in some highly detailed photos, I like the final look of the image more than the simpler routine. I don't bother with the extra work when setting up to print portraits or wedding formals. But when I'm doing some of the landscape work I do, I always give it a try to see if there is a difference that I prefer.
You should look into using PkSharpener, a plug in for photoshop that does a very good job of optimizing sharpening, with a lot less trouble.
The image makes a lot of difference. Something without fine detail - cala lilies - will look great because of the smooth texture of digital, while something with a lot of detail - trees - will emphasize the limited resolution and pixelation.
PKSharpener is fine for RGB images, but it is not optimised to handle Alpha channels and as such, wouldn't be a good choice for the workflow I mentioned. It would be a fine solution for the worflow if you were staying in RGB, but I don't believe I can make the plugin part of an action, and as such I would have to stay the USM.....which can work equally as well as long a s the operator knows what they're doing.
As well, with trees and such targets that require fine detail, it does come down to printing size. The best DSLRs can now easily make 16x24 & 20x30 that have a good amount of detail. You would never have "pixelation" in the image as long as interpolation & sharpening is done right. Increasing the size of digital images with properly done interpolation just results in an image that becomes increasingly softer as output size grows.....pixelation only occurs with poorly done processing.
Powered by vBulletin® Version 4.2.0 Copyright © 2013 vBulletin Solutions, Inc. All rights reserved.