How Does Digital Resizing Affect Line-Art Quality?


-1% ,  Original,   +1%

Open this image then bring it into PhotoShop, or do this experiment yourself...

A 1% resizing of line-art is very noticeable.

Note that any of the "cubic" interpolation tools use the nearest 16 pixels to the one being interpolated.  It will ALWAYS cause grey intermediate pixels where a sharp edge existed.  If you ONLY have line art use "nearest neighbor" interpolation.  Resizing by an  integer factor (1/2x, 1/3x, 2x, 3x...) will also help with these linear interpolation methods.

 

 

How Does Digital Resizing Affect Photo Images?

The debate is whether this degradation is even noticeable in photographic images.  It is said that the lack of sharp edges, lower contrast, and high image resolution make it negligible.  You decide for yourself...

Here is a section of a 3000 x 4500 pixel test image (PDI-Target.jpg).  You can see the degradation with the naked eye at your screen resolution of 72 dpi.  I've enlarged it below so you can better see the difference (or you can run a similar test in Photoshop to see for yourself).  At a print resolution of 300 dpi the image will be about 4x smaller than below, therefore any degradation will be very very tiny and you probably won't be able to see it. (?)

But, if the entire image is degraded like this will a general softening be noticeable? Could it be the difference between a print that just looks good and one that has depth and is almost 3D ?


      original  , +1% resize
.                               (noticeably softer)
 

Here they are blown way up so you can better see the difference.
The resized image is softer appearing.

Note:  I've used Photoshop Bi-cubic resizing, I don't know if the Frontier does it any different or better.
I only looked at a 1% resize, in some cases the Frontier can resize more than 2% (see table below).

 If you still want to entertain the concept of photographic image degradation caused by slight resizing then the following will be of interest.   After all, we worry about film grain, lens quality, tripod shake, and printer resolution while their effect on image quality is probably of the same degree and scale as this resizing degradation.

 

Optimal Image Sizing for Printing to a Fuji Frontier


The Frontier: The Fuji Frontier prints digital files on photographic paper and is a great value for the price range.  Models 350 and 370 print up to 10" x 15" at 300 dpi.  Files should be .TIF, .BMP or .JPG (using minimal compression) and created in the sRGB color space.

The Question: From the above test we see that even a small amount of resizing can degrade an image, especially when it is not necessary.  So, how do we prevent the Frontier from resizing and interpolating your image?  Here is the same question from a different perspective... how do we print an image so that it is dimensionally accurate (i.e. 1 inch on the image = exactly 1 inch on the print)?

The Challenge:  The Frontier typically prints so that the final print is without a border.  To accomplish this the Frontier must actually print the image slightly larger than the paper size. If the paper is 10" wide and your supplied image file is exactly 10" (3000 pixels) wide then your image will be resized/interpolated up between 1% and 2% in order for the final print to be totally borderless.  If you follow the lab's typical advice you will be told to submit a file that is 3000 pixels wide. WRONG...this will cause the Frontier to resize and interpolate your image to make it slightly larger than 10" so that it can print without a border.

NOTE: THIS IS NOT ALWAYS THE CASE.  SOME FRONTIERS USE SLIGHTLY  UNDERSIZED PAPER AND DO NOT RESIZE YOUR IMAGE IN ORDER TO GET A BORDERLESS PRINT !   You can ask the lab what "image size" their Frontier is programmed for, but odds are they won't have a clue as to what you are talking about. The only way to tell for sure is to run a couple 4x6 test prints and look at them under a loupe, see procedure below.

 

 The Solution:  IF YOUR IMAGE IS ALREADY SLIGHTLY BIGGER THAN THE PAPER THEN THE FRONTIER WILL NOT RESIZE YOUR IMAGE.  So, how much larger should your image be?

For print sizes up to 6 inches some (see note above) Frontiers will print 12 pixels beyond each side of the paper size, for larger print sizes they will print 18 pixels beyond the paper size on each side.
Note: The Frontier is programmed in metric so use the metric print sizes as the starting point for all your calculations. Also, rumor has it that being within 10 pixels is apparently 'close enough'. The Frontier will accept this and will not resize the image.

 

The Example:  You want an optimized 8" x12" print (we'll just look at the 8" dimension for this example). 

Bingo, submit an image 2434 pixels wide and it won't be resized .

 

LATEST NOTE:  I SPOKE WITH A REGIONAL FUJI TECHNICIAN AND HE SAID THAT IF YOU GIVE THE INSTRUCTIONS OF "NO CROP" THEN THEY WILL NOT RE-SIZE YOUR IMAGE.  THIS METHOD DOES REQUIRE THEM SWITCHING TO MANUAL MODE WHICH SOME LABS WON'T DO. 
Again, a test print is the only way to know for sure, see procedure below.

 

The List: Some commonly used print sizes.

mm

 inches optimal # pixels % Resizing You Avoid
89 3.5 1074 2.3%
102 4 1228 2.3%
127 5 1524 1.6%
152 6 1818 1%

mm

 inches optimal # pixels % Resizing You Avoid
190 7.5 2280 1.3%
203 8 2434 1.4%
254 10 3036 1.2%
305 12 3638 1.2%
381 15 4536 .8%

 

BOTTOM LINE... Runs some test prints:

  1. Select any test image and using Photoshop place a couple 1-pixel wide dots at 4-5 locations on it.  Save the image.

  2. Now, increase the print size to the number calculated above by adding a black border.  Do this by increasing the 'CANVAS SIZE', not by increasing the 'Image Size'. Save under a different file name.  You may want to add something else to the image (such as an "R") to help differentiate it's print from the original print above).

  3. Have a 4x6 print made of each and look at them under a loupe.  You can very easily tell if they were resized -- the 'dots' will be sharp if there is no resizing and they will be distorted and blurred if resizing occurred (or the machine is totally out of calibration).

Once you discover what you are looking for you will be able to see the difference with the naked eye (gulp).   These are tiny details indeed...but visible is visible.

 

WORKFLOW FOR BEST QUALITY:

The goal here is to only resize only ONCE to get your desired print size!

  1. Scan at the exact scanner hardware resolution.  Anything else causes the scanner to interpolate using it's own internal (unknown) technique.  Their internal routine is possibly as good as but not liekly better than available software resizing solutions.  See the post from Dave Martindale in rec.photo.digital on 8/19/2002, "Subject: Re: Resolution - digital camera, scanner" (or any of his
    other posts) for more detailed explanations.
     

  2. Resize to the exact hardware resolution of the printer, take into account your resizing test results above.  300dpi for Frontier and 720dpi for Epson print drivers (see step 2 at http://www.ddisoftware.com/qimage/quality/ for details).  For "typical" photo images use one of the cubic resizing filters like Lanczos or PhotoShop Bi-cubic (Don't use IRFANView, it has a bug that causes ALL it's resizing methods to give the exact same results!).  For pixilated or jpeg artifact loaded images use B-Spline.  For huge upsampling use Genuine Fractals.
     

  3. If you are printing through a lab always instruct them "NO CROP, NO RESIZE, NO COLOR ADJUSTMENT".

 

Let me know the results of your testing, drop me a note:  Jack

This information is gathered from personal experience and several resources on the web. The folks at Color Services, CA. were extremely knowledgeable and helpful.  The table on their site provides optimal image sizes for many print sizes: http://www.sbphotonews.com/File2Pri.htm

Roger Clark has excellent discussions on digital imaging on his site.
 

 

Some of my other pages:

 

Hit Counter