Curt Y wrote:
Here is a long thread on upsampling/downsampling. Basically automatic mode is using smoother for upsampling and sharper for downsampling.
It's not clear to me from reading that entire thread whether they are talking about resampling images using Image Size, or using options under View (Zoom In, Zoom Out, Fit to Screen, etc.)
It seems they are mostly discussing the former, whereas I am referring only to the latter (which is why I used the term "on-the-fly" in the Subject of this post).
If you are just discussing changing viewing on the screen no changes will be made to the image. Also, if you change size and resampling is off no changes are made. It is only when you resample the image to either add or subtract pixels that some form of pixel modification is performed.
If you are just talking about scaling then pixels are not affected. Here is another reference that discusses the subject http://www.scantips.com/lights/resize.html
Curt Y wrote:
If you are just discussing changing viewing on the screen no changes will be made to the image.
The image file will not be changed but something must be done to display a smaller image on the screen. If the original pixel dimensions are 6305 x 4181, and the image is displayed at 1920 x 1080 pixels, somehow Photoshop has to reduce the number of pixels it uses for the display. If I understand the OP, he wants to know whether PS uses a Bicubic algorithm to do this ("on the fly"). My guess is that PS uses the same criteria it would use when resampling (Smoother for upsizing, Sharper for downsizing).
Rafael Aviles wrote:
. . . If I understand the OP, he wants to know whether PS uses a Bicubic algorithm to do this ("on the fly"). . . .
Yes, Rafael, this is the crux of my question.
One thing for sure, Apple's Preview and Photoshop do something different simply to display the same image at the same size. So the issue becomes, when analyzing scans of images on film, prior to any editing, what am I really seeing?
Yesterday, I showed this mystery to a trainer at the Apple Store (on my iMac, which I brought into the store.) He believes that I am seeing a long standing issue with how Preview renders high resolution TIFF images. He found this forum discussion for me:
Preview tiff distortion
I think he's right. Therefore, I am no longer concerned about what Photoshop is doing when it displays the images.
Just for reference...
Photoshop's document resampling settings have nothing to do with down-zoomed preview displays.
That process is done partly in Photoshop and partly in the video card, where (as I understand it) a modified bi-linear algorithm is used to resample a pre-sized cache image to provide you the exact zoom you request.
Assuming you're using GPU acceleration, which is the default, as you have noticed it's smoother than the resampling that Photoshop does during non-GPU-accelerated operation, but it's not exactly equivalent to any of the Photoshop-offered resampling modes.
Bilinear resampling represents a balance between quality and speed in the GPU. Since GPUs are capable of more graphics processing speed than the CPU, a smoother job can be done in a tiny fraction of a second than the old CPU-specific resampling Photoshop did in the days before GPUs.
But it still uses those algorithms to generate downsized cache images. Under the covers the GPU does not always actually resample the full-sized original image, but (depending on the zoom level) it is given a downsized version. There is a setting in your Performance preferences called Cache Levels that defines how many successively downsized images you want automatically generated internally from which to drive the display you see at various zoom levels. That also is a measure intended to balance speed against quality, the idea being to give you an interactive experience even with a giant image. You can, if you want, set Cache Levels to 1, then the preview image really is always generated from your entire full-sized document. But watch out, that can expose bugs and it will be SLOW even on a rompin' stompin' computer.
So you see, there's FAR more going on behind the scenes than you'd think, and it's not always perfect... For example, if you zoom to exactly 50%, you may notice it looks slightly sharper than at 51% or 49%.
Thank you, Noel. This is fascinating information.
Right now, my Cache Levels is set to 4. I may keep it there, but it's going to be interesting to experiment.
If you do try Cache Levels 1, and experience a crash, try setting the Advanced GPU mode to "Basic" Drawing mode, that can help with stability.