Resampling Images by Jeff Ball

Let's see, a 2184x1472 chip is capable of producing a print size of 9" x 6"  at 240 dpi.  About 2 years ago, that was just gibberish to me.  Not that I have a handle on it, but I think I do have some understanding today.  It seems that with all of our technology, most of us still want a print on paper to hang on the wall.  Recently, I began a study of the best way to maximize my  ST10E chip (2184x1472) for printing (I am always open to suggestions.)

Many printers / print shops recommend at least 240 dpi (dots per inch) resolution for a quality print.  The calculation above shows the native print size at 240 dpi for my ST10 chip.  If I want a larger print, I must resample the image.  How far can one resample?  I don't know.  I have found that 12"x8" prints look fine when I have an image with good signal to noise with the ST10 chip.  Is the need for resampling an exclusive issue to CCD?  Perhaps not.  Here are some numbers (this is by no means an exhaustive list, just some common examples.) 

 

Image recorder Native size Print size at 240 dpi
ST10 CCD 2184 x 1472 pixels 9"x6"
35 mm film scanned at 2400 ppi 3304x2348 13.7"x9.7"
6x7 120 format film scanned at 3000 ppi (180mb file with some cropping) 6418x4900 26.7"x20.4"

Resampling may be utilized anytime one needs a print larger than the native format allows at 240 dpi or at an even higher dpi.  Cropping a full frame film image or scanning at a lower pixels per inch may also result in the need for resampling to achieve the desired print output.  However, scanning at higher ppi will most likely reduce the need for resampling of film images.  When scanning at higher ppi your ability to enlarge will be limited by the resolution of the image and the grain in the emulsion. 

No question, resampling is more of an issue with the less than 35mm full frame CCD and CMOS chips than it is with film.  The "real" resolution of the digital file will determine the amount of enlargement.  A high signal to noise image can withstand a large amount of sharpening through various operations.  Prints as large as 30"x20" are being made by some astroimagers from a 2184x1472 size chip. 

In this review I look at 3 common software programs with regards to how well they resampled an ST10 image.    The programs reviewed were:

1.  Adobe Photoshop 5.0 (retail price for version 7.0  ~ $649) All prices as of January 2003

2. Picture Window Pro 3.1 (retail price ~$89.95)

3. S-spline 2.2 (retail price ~ $69.99)

There is another popular image resampling program called Genuine Fractals.  I do not own this program.  It sells for ~ $139.95 and there is a lite version for ~$49.95.

Photoshop and Picture Window are complete image enhancement programs with comprehensive capabilities.  S-Spline and Genuine Fractals are focused programs in that their main operation is to resample images for larger output sizes.



Below are four examples of an ST10 resampled image to achieve a 12"x8" print at 240 dpi.  This means that the image has been enlarged by 77%.  I have cropped the image to focus on the most revealing part and to allow a 100% view at web resolution.  This area of the image should give an indication as to how well the resampling operation performs in regards to sharpening stars, nebulosity, or increasing the noise in the image.  The resampling operations are all bicubic with the exception of the final S-Spline image.  I looked at whether or not the 3 software programs performed the bicubic operation similarly.  You be the judge.  Picture Window does offer the user an option to adjust the amount of sharpening that takes place during the resampling operation.  I used a value of 50% for the example below.  For a more convenient view of three of the images, please click here.

For a full frame look at the image and more information, please click here.
 





A bicubic resampling with Photoshop 5.0
 





 
A bicubic resampling with S-Spline 2
 





A bicubic resampling with Picture Window Pro with a 50% sharpening.  PW Pro has an adjustable sharpening tool to apply to the resampled image.  This may prove to be very useful with experimentation.
 





S-spline resampled image using S-spline's proprietary operation. 
 



Conclusions:

Which one do you think is best?  Any differences?  Well, I think S-spline is best.  Maybe I am just trying to justify my purchase. Still not sure?   Here is a super comparo enlargement over 1100%!  Two images.  The left is S-spline and the right is bicubic in Photoshop.  The image is large enough for a 240 dpi print at 30"x 20".  I think if you look at the lines of contrast between the nebulosity and the dark nebula they are much sharper.  The stars are definitely sharper than with any other resampling operation.  At first, I thought the stars were too sharp, but on prints they look great.  I actually think the S-spline image is a little less noisy in the shadows.  It is very difficult to see a major difference.  Right now, I am quite pleased with the 12"x8" prints I am getting with S-spline on my single frame CCD image. 

Want more information?  See the review of S-Spline at Adirondack Video featuring some terrific images by Rob Gendler at this link.  Some of the examples at Adirondack show a 225% enlargement.  A review of Genuine Fractals may be found here with a second review here.

Please let me know if any information on this page is incorrect.  Thanks for taking the time to read and best regards.

Jeff Ball

Home