OK, I'm not sure if the question I have has a direct answer or not--but I'm going to give it a shot anyway.

Let's say that I have a 4 x 6 print. The client does not have the negative. They want the photo retouched and they want the new print to be 8 x 10 in size. Is there a way to know whether or not your scan contains enough information to make this enlargement? I mean, I know you need a certain amount of pixels per inch and such, but how do you know whether or not those pixels contain useful information? Is there a point when you should have your scans outsourced to a more capable scanner? I would hate to spend hours working on a image that looks unprofessional once it's printed out.

I've always read that you should scan for the largest possible size that you think you will need. My thinking is that if I figure out how to make a scan that will produce enough information for a decent 8 x 10 then I've got all my bases covered. Is this logical thinking?

Amanda

Amanda,

I'm *real* good at trying to answer questions that I'm not particularly qualified to do :) . But it seems to me that if the 4 X 6 is a good print, you should not have a problem enlarging it to 8 X 10, especially if you do not exceed the optical resolution of the scanner. But *anything* in the original that is not good (focus for example), will become more apparent with the enlargement. One thing you have going in your favor is the fact that the 8 X 10 will not have the same viewing distance as the 4 X 6. Of course you can always run into that customer who is a major pain, and might not even be satisfied with an 8 X 10 from a neg. Before you take this post as being fact, you should probably do a test on one of your own prints to see just how far you can go with it. I think you'll be happily surprised. Just my two pennies, and I'm sure if there are other opinions by someone more qualified, they'll put me in my place. :(

Ed

enlarging or shrinking digitally relies on your graphic programs ability to either 'add in' or 'take out' pixels.

Based on that, any resizing of an image digitally will result in poorer quality. Fact of 'digital' life I am afraid. The basic failure in the digital world, is that we are relying on the program to interpolate the existing pixels. In other words, we ask the program to either add in , or take out pixels. If you know that you are going to be asked to either enlarge/shrink an image, providing the program with the maximum amount of pixels is advisable.

digital 101 :) ~Vp~

Vp,

I'm not sure I understand exactly what you're saying. If she scans the image at say, 600 spi, then reduces it to 300 ppi in Photoshop to make an enlargement (no resampling), I don't see how you could consider that interpolation. Am I missing something here? Quite possibly I just don't understand it correctly, but that's not exactly new for me. :)

Ed

I was always under the impression that any form of resizing an origial digitally asks the software to 'guess' what it throws out or what it adds in depending on whether your shrinking or enlarging.

I am not 100% certain. Maybe if it enlarges in mathematical increments it can do a perfect job. I would like to know, as I am not an expert by any means.

~Vp~

i have needed to know this very thing. (thanks amanda:wavey: feels like you been readin my mail) note to self: if you don't understand something, chances are somebody else doesn't either, so ask.

ed, if i do what you describe, @ what point to you change the res down to 300 - before or after you change the image size? does it matter?

Akj, Scanning at around 300 to 400 dpi will give you enough data to make a good 8x10 blow up from a 4x6 size. If you have around a 7 to 10 mb file size, after resampling, you should be OK. One way around the hassle of large file sizes and their demands on memory, would be to check out the program GENUINE FRACTALS which allows you to resize with minimal loss of detail and artifact production. I discovered last year while living through the nightmare of doing 5 schools worth of Sports Pics that a 5x7 or a 4x6 scanned at 300dpi could be brought up to an 8x10 in Photoshop 5.5 using bicubic resampling with great results. Usually these files were in the 7 to 10 mb range and printed at 200 dpi on heavyweight matte paper. I generally look at the file size to make an estimate of whether or not there is enough "density", image data, to enlarge without creating a pixelated mess. There are simple calculator programs, usually freebies out there which can greatly assist in this type of determination. Visit the DIGITAL LIGHT AND COLOR web site(I think Doug put up a link over in the Salon under the thread PICTURE WINDOW) and look for the download entitled SCANCALC. Its an elegant and simple little program which can be of great help. Hope this helps, Tom

Ed and vogonpoet, As I understand it, most if not all of the resampling algorithms function by averaging pixel values then either adding to or subtracting from the image. EXCEPT for the Fractals algorithm which uses complex math to "build" the image resample "order" in increments as vogonpoet mentioned.Because of the way in which the calculations are made and implemented, a Fractal resampled image can be enlarged or shrunk by huge percentages without loosing very much clarity or detail, allowing for a low density scan to be made and then resampled as needed for different size applications without the need to make multiple scans at different resolutions. However any manipulation of a digital image will result in some degree of degradation,even just opening and saving it. HOWEVER this is usually so slight that it is not a problem visually although it can be "seen" by such means as the histogram. Hope I got all that right or its a tar and feather party with me as guest of honor to commence immediatly after I press the POST button! Tom

:D

exactly! ~Vp~

For most printers (and I'm talking inkjet here) 150dpi is about the lowest rez where people won't start seeing jaggies, squares, pixels, etc.

So take your original dimensions, multiply by the resolution (not counting software interpolation or enhancement) and you get your total pixels tall by wide.

Then calculate the enlargement dimensions and divide that into your total pixels. If it drops below 150 you'll probably be in trouble (more is better, and again we're talking about real optical resolution here).

So, if you have a 4 x 6 inch print, scanned at 600dpi, thats 2400 x 3600 pixels. You're wanting an 8 x 10 (we'll ignore proportions for ease of math's sake here). So 2400/8 = 300 and 3600/10 = 360. So whichever proportion you choose should be fine.

Did I understand the question correctly? (or am I making things worse?)

does that mean when you go to the image size box and put in your new larger dimensions, you leave the res. box unchanged? no touchy?

That's all interpolation, so really all it does is slow things down. No harm, no foul.

What I do is finish my restore, save, make a copy, and then resize that copy physically in Photoshop to the paper size. This allows me to set the canvas size to the final print size and compose the image appropriately. This is also usually the first and only place I'll use USM.

I make my prints, then save as a flattened compressed TIF, in case they reorder.

Doug,

I think I understood the question the same way you did. Would it be accurate to say that all pixels (not interpolated) should be thought of as having useable information, but that the quality of the enlargement would be directly related to the quality of the original, when enough optical resolution is used for enlargement? *Unsubstantial* losses such as Tom described can be disregarded (such as losses created from opening and closing a file).

Isn't interpolation when you increase the number of pixels in the image? If so, lowering the resolution (without resampling) would increase the *size* of the pixels, but not the *number* of pixels, thereby creating a larger print size without interpolation (although *unsubstantial* minor degradation might result). If I'm not clear on this, somebody needs to drill this into my thick head.

Tom,

I've heard about Genuine Fractals before. Some people say that you can *greatly* increase print size without noticeable loss when using this program. Just how good is it?

Ed

Ed,

I'm not sure if I follow you or not, but I've got the drill ready...

It sounds like you're basically just saying increasing the size without resampling lowers the resolution, which is true. Less pixels per inch because the same pixels are spread further apart.

Maybe I don't understand the concept correctly. Are you saying that there is void space between the pixels, or are the pixels becoming larger? I guess I'm not making it clear what my understanding (which may not be correct)of the process is. I'm thinking that when the same number of pixels covers more area (inches), the pixels actually become larger, and still border each other with no space between.

I think it was Debbie who suggested that if she didn't understand something, there are probably others who also don't understand it. So, rather than look stupid, :( I'll go along with that concept.

Ed

The pixels touch. No void. It's just a numerical statement of the value at a given coordinate.

How many coordinates are used determines the resolution. If you enlarge using the same number of coordinates, that numerical value has to be spread to cover a larger area.

Interpolation looks at these 'coordinate values' and guesses what should fall between. The varying theories as to HOW it should guess make up all the different resampling schemes.

There's very little corelation between the number of pixels per inch in a file and the dots per inch a printer is capable of laying down. To ease the concept, assume the resolution of the printer is infinite and perfect (it isn't, of course, but we're talking theory here).

Imagine a 'pixelated' image. Big squares, jaggy diagonals. That's the coordinate values being spread to cover larger areas.

Hi All--it looks like I've got everyone thinking here. :) Thanks for all your input.

A question for you Tom:

Originally posted by thomasgeorge
Akj, Scanning at around 300 to 400 dpi will give you enough data to make a good 8x10 blow up from a 4x6 size. If you have around a 7 to 10 mb file size, after resampling, you should be OK.

Did you mean to say resampling? Or scaling? If you change the image size (with resample box unchecked) this will change the resolution but not the actual number of pixels (otherwise known as scaling). So shouldn't the file size stay the same before and after you've changed the dimensions? If you have the choice between resampling an image and scaling it--isn't it best to scale?

So, to touch on what Doug said--if I scan in an image at let's say 400 pixels--I scale the size up--and as long as the resolution is 150-200 dpi, I should be OK? And, like Ed said, the quality of the enlargement will depend entirely on the quality of the original. So, I guess the only way to judge whether an enlargement is satisfactory enough is to simply print it out and see? Should I not worry so much about "scan quality"?

:( Maybe I may be making this topic more complicated than necessary.

Amanda

Doug,

I think what you are saying is exactly what I had in mind to begin with.

Amanda,

You are not making this thread too complicated. I think it got a lot of responses because of the way posts were worded and/or interpreted. (Hope this doesn't start another one) :)

Ed

Akj, You are right. Poor choice of words on my part. I was refering to upsampling which can be either fixed or variable resolution type. Sorry! With fixed resolution type the file size increases. With variable type the the file size,actually the number of pixels, is unchanged. Tom

Ed, the program Genuine Fractals from Altimira is useful if you do alot of enlarging or compositing work as you can scan or otherwise import an image at a relatively small size then upsample or down sample it as your needs dictate without having to have multiple images at different sizes stored on your hard drive eating up valuable storage space. Unless you do lots of that sort of stuff where the work must undergo some pretty radical changes, it probably isnt a necessary "plug in" to aquire. I only use it a couple of times a month if even that but when you need it its nice to have. It works well, but keep in mind that VIEWING DISTANCE enters into the equation as well. An enlarged image viewed from say 2 feet is going to look better than the same one viewed from 6 inches. The bigger the finished image the more "fuzz etc " there may be, and Fractals cant compensate for a borderline original image.If its fuzzy or otherwise less than perfect, so it will be after fractals finishes its work. Tom

Thanks Tom. It's probably not something that I'll get, but like I said before, I've heard about it, and it sounded like it might be something interesting. I don't remember where I heard about it, but it sounded kind of "magic like".

Ed

Ed , Alot of the programs out there especially the ones dealing with improving image quality when importing or manupilating size sound real good, and to be fair most do have some value. Trouble is none of them will do much without the user paying attention to detail and using common sense when implementing a course of action and most of the programs with the exception of some of the scanner programs, like the one you use, are pretty narrow in what they do. Its easy to drop a bundle and discover too late that what you thought you were getting isnt exactly the absolutely necessary must have or perish thing you initially thought it was. I know from sad experience, of course a hundred years or so ago I'd probably have been the first yokel to crowd up to the Patent medicine sellers wagon with a buck in my fist. Tom