Bart_van_der_Wolf
pro member
Hi folks,
I just wanted to share some of my findings in a quest for the best upsampling method. Some of us routinely output large format output, others may be confronted with an image that, e.g. due to cropping, just lacks the number of pixels needed for sharp looking output. I'm not addressing sharpening issues as such, although the may become part of an upsampling strategy.
Upsampling can also be used to promote an image that (almost) satisfies the 300 or 360 PPI output required(*) by our printer drivers, to a 600 or 720 PPI image, which will make a difference with smoothness of gradients and sharper output.
(*) Note: Printer drivers will interpolate all input data to allow the optimized use of their native resolutions, 300 or 600 for e.g. Canon and HP printers, 360 or 720 PPI (720 PPI requires setting the finest detail option) for Epson printers. The resampling algorithms are relatively simple compared to what can be done with image processing software.
Here is a sample (crop from an) image, taken with my 1Ds3, straight from the Raw converter, without sharpening:
Here are the 400% enlargement results of 2 different upsampling methods, Photoshop's BiCubic Smoother and Photozoom Pro's S-Spline Max.
And here are the results after 800% enlargement with BiCubic and S-Spline Max.
These sample crops are best previewed at something like 25% zoom for an approximate impression of how it would look when printed, but feel free to actually make a print to see what it really does as printed output. One can also add one's favorite output sharpening method for a more realistic comparison.
Do understand that for these last (800%) examples one is looking close-up at a fragment of 3.1 to 3.8 metres wide output (depends on printer PPI setting) from a single unsharpened 1Ds3 (36x24mm) file, which is of course not as detailed as would be possible from a larger file. However, one is not always able to shoot the scene with a slower heavier camera, or use the result from stitching. Sharpening will add some more punch to the results.
Resampling is no substitute for more real pixels, unless we don't have more pixels.
Cheers,
Bart
I just wanted to share some of my findings in a quest for the best upsampling method. Some of us routinely output large format output, others may be confronted with an image that, e.g. due to cropping, just lacks the number of pixels needed for sharp looking output. I'm not addressing sharpening issues as such, although the may become part of an upsampling strategy.
Upsampling can also be used to promote an image that (almost) satisfies the 300 or 360 PPI output required(*) by our printer drivers, to a 600 or 720 PPI image, which will make a difference with smoothness of gradients and sharper output.
(*) Note: Printer drivers will interpolate all input data to allow the optimized use of their native resolutions, 300 or 600 for e.g. Canon and HP printers, 360 or 720 PPI (720 PPI requires setting the finest detail option) for Epson printers. The resampling algorithms are relatively simple compared to what can be done with image processing software.
Here is a sample (crop from an) image, taken with my 1Ds3, straight from the Raw converter, without sharpening:
Here are the 400% enlargement results of 2 different upsampling methods, Photoshop's BiCubic Smoother and Photozoom Pro's S-Spline Max.
And here are the results after 800% enlargement with BiCubic and S-Spline Max.
These sample crops are best previewed at something like 25% zoom for an approximate impression of how it would look when printed, but feel free to actually make a print to see what it really does as printed output. One can also add one's favorite output sharpening method for a more realistic comparison.
Do understand that for these last (800%) examples one is looking close-up at a fragment of 3.1 to 3.8 metres wide output (depends on printer PPI setting) from a single unsharpened 1Ds3 (36x24mm) file, which is of course not as detailed as would be possible from a larger file. However, one is not always able to shoot the scene with a slower heavier camera, or use the result from stitching. Sharpening will add some more punch to the results.
Resampling is no substitute for more real pixels, unless we don't have more pixels.
Cheers,
Bart
