|
| Author | Thread |
|
|
04/30/2004 07:25:16 AM · #1 |
Would I be right in saying that interpolation is an incredible advantage of digital photography?
I've heard friends say that if you enlarge film prints - the picture only goes grainy, but with digital you get that nasty pixellation.
I shoot with a 4mp camera. In photoshop I can, for example, triple the image size. Photoshop seems to do an amazing job of guessing how to fill in the gaps - often it seems to guess accurate detail out of nowhere.
Could I draw potentially huge prints from my 4mp original pictures?
What are the limits?
Why haven't I heard more about this as an amazing feature of digital photography?
Are there any disadvantages of interpolating images up? (apart from increased file size / unmanagability).
At the very least it seems that I could eliminate the appearance of pixellation going up way past A4 size with pretty decent results. Far higher than even a decent 35mm film image, no?
Since you can do this with any image editing software - why does Fuji trumpet its CCD interpolation technology as groundbreaking?
Thanks
And thank you also to everyone who has answered my previous queries - I really appreciate your help. |
|
|
|
04/30/2004 07:28:18 AM · #2 |
It's generally recommended not to increase the original size of the image by more than 50% as quality does start to drop at that point.
It's also suggested that, if doing this in Photoshop, you increase in increments of 5% until you reach the required final size, rather than in one step.
There are applications available that do this for you.
I think Setz knows more about this than I do, as do many other members here.
I'm sure someone will have a better answer for you.
|
|
|
|
04/30/2004 08:08:02 AM · #3 |
A good example is roddy's Dead Vespa if you open the 150k image in Photoshop and interpolate it up by about 10 times you can draw an unbelievably detailed image out - making the logos/signs legible when zoomed in. The potential with a 2/3MB image is awesome.
How on earth does Photoshop do it? |
|
|
|
04/30/2004 08:08:24 AM · #4 |
You are both right and wrong in saying that interpolation is an advantage of digital photography. Any image that you can put into an image editor can be interpolated either up or down in size.
Yes you can create huge prints from a four mega pixel camera, but you will lose quality. There are many different algorithms for interpolation, but the main ones are as follows: --
bi-cubic interpolation -- cubic curves are used in an attempt to best synthesise the data between two points.
Nearest neighbour in interpolation -- a pixel's colour values of copied into more pixels in order to enlarge on image.
Linear interpolation -- the difference in colour values between two pixels is taken as though there is a linear progression between the two pixels the interpolated pixel values are read from the straight line.
I hope that was clear, but it probably wasn't.
The most major disadvantage to interpolation is that you cannot create information that was not there in the first place. That means that you are only guessing at what information may have been there. Even the most advanced heuristic's algorithms cannot create information that was not there in the first place.
The limits are what you are prepared to put up with in terms of quality. That is to say that you will decrease the sharpness of a picture if you interpolate it. Also when you are considering what quality you are prepared to put up with do not forget that your monitor will have a resolution of around about 70 pixels per inch, your print resolution will almost always be different to this. Assuming that you use 300 pixels per inch printer resolution then to get an idea of the quality of the final print you should view the image at about 25% in your image editor.
There are programs that claim to be incredibly good at interpolation, however you will still lose quality. Two of these programmes are as follows: --
pxl smart scale
genuine fractals
To answer your final question about Fuji, the reason that they claim that their CCD technology is groundbreaking is because of the structure of the CCD rather than the interpolation that they use. Consider their CCD to be like any other that is in any camera at the moment, but to have been rotated by 45°. This means that effectively two horizontal pixels are further apart than in a normal CCD, the same goes for vertical pixels. However as you can have a similar quantity of actual photo sites as in a normal CCD you can use this angled CCD layout to interpolate more accurately what information would have been between two photo sites. In actual fact whilst images are amongst the best that you can get at any given resolution at the moment, they still lose quality when being interpolated. If you should doubt that, then compare the latest Fuji to the 1Ds from Canon. The Fuji claims to be 12 mega pixels, the Canon nearly is.
I hope that you find this information helpful. If you have any further questions and sure that either myself or somebody else will be able to help you. |
|
|
|
04/30/2004 08:57:37 AM · #5 |
Originally posted by budokan:
Could I draw potentially huge prints from my 4mp original pictures? |
The answer is yes.
To give you an idea, look here for some sample enlargements done from a 3MP Canon D30. |
|
|
|
04/30/2004 09:36:59 AM · #6 |
Originally posted by EddyG:
To give you an idea, look here for some sample enlargements done from a 3MP Canon D30. |
Amazing!
|
|
|
|
04/30/2004 09:44:05 AM · #7 |
Originally posted by sn4psh07:
The limits are what you are prepared to put up with in terms of quality. That is to say that you will decrease the sharpness of a picture if you interpolate it. Also when you are considering what quality you are prepared to put up with do not forget that your monitor will have a resolution of around about 70 pixels per inch, your print resolution will almost always be different to this. Assuming that you use 300 pixels per inch printer resolution then to get an idea of the quality of the final print you should view the image at about 25% in your image editor. |
The 'dots' on your screen are not the same size as a printed image. Also, the screen emits light, as opposed to print so the appearance is very different. Most everyone agrees to shapren an image for screen viewing, but no or much less sharpening is needed/recomended for printing.
Originally posted by sn4psh07:
In actual fact whilst images are amongst the best that you can get at any given resolution at the moment, they still lose quality when being interpolated. If you should doubt that, then compare the latest Fuji to the 1Ds from Canon. The Fuji claims to be 12 mega pixels, the Canon nearly is. |
Agreed, but don't compare apples and oranges. It is a true 6mp camera (has 6 million sensor sites) but can produce an image that is in overall pixel count (size) about the same as a 12mp camera. Most seem to kind of split the diff quality wise and say it is the same as an a 8mp camera in the end results.
Not to confuse things, but the new S20Pro does things the same way, but at each of the 3 millions sites are 3 million more that capture the lighter pixels, extending the dynamic range of the camera. So it has a true 6 million sensor sites, but those generate a file of 3 million pixels, just cover a larger range of stops.
For more on CCD technology, check this site
here.
See Here.
As to how big a print you can make - depends upon viewing distance. A normal print is 200 to 300 dpi. Drop that to 100 and you double/triple the print size, but will have to stand back 3 feet maybe to avoid the pixelization effect. Roadside billboards are printed at 2 to 6 dpi, but from the distance you see them they look fine. Newspaper photos are 150 dpi (screened).
Take a part of a pic and blow it up to the equivalent 30x40 inch or what not and have it printed as a 4x6 inch at Walmart for 26 cents. A cheap test to see how big you can go and what the end result looks like.
|
|
|
|
04/30/2004 09:48:16 AM · #8 |
For the images linked above, of course it could be debated if you could see how good the quality is in such small pictures. The images are much smaller than the 3MP prints. So they could look like total crap and still appear as amazing as they do. Just a thought.
Apart from that, it's clear that when enlarging digital files over their "natural" size, they clearly win against analog enlargements. The reason is that with analog enlargement, you also enlarge the film grain, edges get soft, etc. But the filters that are normally used to interpolate (especially the bicubic filters, and there are even better ones) will leave edges sharp and will not enlarge the underlying image structure (pixels). |
|
|
|
04/30/2004 12:43:46 PM · #9 |
Originally posted by bestagents:
The 'dots' on your screen are not the same size as a printed image. Also, the screen emits light, as opposed to print so the appearance is very different. Most everyone agrees to shapren an image for screen viewing, but no or much less sharpening is needed/recomended for printing.
Agreed, but don't compare apples and oranges. It is a true 6mp camera (has 6 million sensor sites) but can produce an image that is in overall pixel count (size) about the same as a 12mp camera. Most seem to kind of split the diff quality wise and say it is the same as an a 8mp camera in the end results.
Not to confuse things, but the new S20Pro does things the same way, but at each of the 3 millions sites are 3 million more that capture the lighter pixels, extending the dynamic range of the camera. So it has a true 6 million sensor sites, but those generate a file of 3 million pixels, just cover a larger range of stops.
For more on CCD technology, check this site
here.
See Here.
As to how big a print you can make - depends upon viewing distance. A normal print is 200 to 300 dpi. Drop that to 100 and you double/triple the print size, but will have to stand back 3 feet maybe to avoid the pixelization effect. Roadside billboards are printed at 2 to 6 dpi, but from the distance you see them they look fine. Newspaper photos are 150 dpi (screened).
Take a part of a pic and blow it up to the equivalent 30x40 inch or what not and have it printed as a 4x6 inch at Walmart for 26 cents. A cheap test to see how big you can go and what the end result looks like. |
Fuji compared it to 12 MP, and I was trying to make a point about interpolation hence my comparison! Also the new super CCD you speak of still has 3 million photo sites, it has 6 million photo diodes :)
A dot is a dot, my point was the picture will look worse on a monitor - and it will! Also for printing, if you do home prints on an inkjet then you need to sharpen more as the ink flows after hitting the paper :)
If you don't work in the print industry you won't tell the diff between 150 dpi and 300 dpi. It's like I said what quality will you put up with :) |
|
|
|
05/01/2004 06:56:18 AM · #10 |
I understand that enlarging a photo - whether interpolated or not, will reduce quality. What I was getting at was:
When enlarging film - there is no interpolation, right? So however big you go, you are still working with the same information that was captured - and simply blowing it up. Is this correct?
When enlarging digital pictures, however, you are doing the same thing but filling in the gaps and smoothing over the edges - thus adding more information (albeit not genuine information, just a good guess at it)
So my specific question is: isn't this a massive advantage OVER FILM, since you can enlarge a 4mp digital image many many more times than a 35mm film image? |
|
|
|
05/01/2004 08:26:39 AM · #11 |
Originally posted by budokan:
I understand that enlarging a photo - whether interpolated or not, will reduce quality. What I was getting at was:
When enlarging film - there is no interpolation, right? So however big you go, you are still working with the same information that was captured - and simply blowing it up. Is this correct? |
Yes, absolutely.
Originally posted by budokan:
When enlarging digital pictures, however, you are doing the same thing but filling in the gaps and smoothing over the edges - thus adding more information (albeit not genuine information, just a good guess at it) |
Yes again. No new information, just a guess at what should be between. Things get "blurrier".
Originally posted by budokan:
So my specific question is: isn't this a massive advantage OVER FILM, since you can enlarge a 4mp digital image many many more times than a 35mm film image? |
Not really. When you do a film enlargement, the enlerger lens is doing approximately the same thing as interpolation. You don't see "pixelation", you see smooth blending.
Where digital does have an advantage is with the ability to use more sophisticated interpolation methods, and to apply special mathematical methods like USM, deconvolution (e.g. FocusMagic, DxO), and noise reduction. But you can do this with film as well by scanning to digital...
|
|
|
|
05/01/2004 10:20:10 AM · #12 |
When you enlarge from film, the original "pixel resolution" of the negative is between 4-8000 ppi (the size of the grain), so that although there's no interpolation, the original pixels can be enlarged and spread out quite a bit before the grain or "pixelization" becomes visible.
If you take a film with a grain approximately equal to 4000 ppi (around ISO 200 speed?) you'd be able to enlarge a 35mm frame up to a 13x20 print and still have the equivalent of 300 ppi.
Also, I'm going to guess that optical enlargement may cause some blurring/blending of the film grains themselves, without causing visible blurriness in the larger shapes they define. |
|
|
|
Current Server Time: 04/19/2024 06:23:32 AM  |
Home -
Challenges -
Community -
League -
Photos -
Cameras -
Lenses -
Learn -
Prints! -
Help -
Terms of Use -
Privacy -
Top ^
DPChallenge, and website content and design, Copyright © 2001-2024 Challenging Technologies, LLC.
All digital photo copyrights belong to the photographers and may not be used without permission.
Current Server Time: 04/19/2024 06:23:32 AM EDT.
|
