upscaling images in pdf
- Thread starter mazengh
- Start date
Stephen Marsh
Well-known member
What is the expectation from doing this? …
Something a little more blurred/softer than the original, taking up more disk space and RIP time for no practical gain and benefit?
Or are you expecting the image to actually look better than the original?
Are these raster logos or raster photo content?
Enfocus PitStop Pro 13 has two bicubic resample options, I have not compared them to Photoshop though (however my guess is that it does not really matter).
GIGO rules.
Stephen Marsh
Something a little more blurred/softer than the original, taking up more disk space and RIP time for no practical gain and benefit?
Or are you expecting the image to actually look better than the original?
Are these raster logos or raster photo content?
Enfocus PitStop Pro 13 has two bicubic resample options, I have not compared them to Photoshop though (however my guess is that it does not really matter).
GIGO rules.
Stephen Marsh
Last edited:
gordo
Well-known member
I have to somewhat disagree with you. You can usually upsample images by a factor of 2x without losing print image quality because the halftone screen will typically clobber the detail anyway. What upscaling the image will do is help get rid of staircasing on image detail like straight edges vs just upsizing the image without resampling..
Just for fun I took an image and resampled by a factor of 2x in PShop:
The method used from top to bottom:
Original image
Scaled without resampling
Bilinear
Nearest Neighbor
Bicubic (Smooth Gradients)
Bicubic Sharper
Bicubic Smoother (the OP's choice)
Preserve Details
Note the difference in the edge of the iris - the scaled without resampling is clearly pixelated vs the resampled versions (except the nearest neighbor one).
Last edited:
The PDF specification provides for an interpolation entry in image dictionaries. When set, (it's boolean - either on or off), it instructs the output device to interpolate the pixels at device resolution. I don't believe compliance is considered to be mandatory, and there is no method of interpolation suggested. Acrobat (and our RIP, which is the Adobe PDF Print Engine) uses bilinear interpolation on images that have the property set.
The good news is that it's essentially cost-free in terms of file data since you're not adding a bunch of extra pixels, and you can do it with Acrobat Professional since version 8 (earlier versions will show interpolation but provide no way of turning it on). I believe Enfocus Pitstop can check for it, but cannot turn it on or off.
It is possible there are devices/readers that would honor interpolation for 8 bit data, but not 16 bit, or indexed, or ICC tagged, etc., so you would want to test heavily if this is very important.
I tend to agree with Stephen. Interpolation turns sharp pixel edges into blurry gradients, and I personally believe that the low-resolution look is more forgivable than the I-took-a-blurry-out-of-focus-image-and-slapped-it-on-my-newsletter look.
Gordo, I think you might have a point for limited resolution increases - particularly for the fancier algorithms that are almost interpolation and sharpening combined.
To change it with Acrobat, you need to create a preflight with a single "fixup." I've attached a screen shot of the fixup that gets it done.
The good news is that it's essentially cost-free in terms of file data since you're not adding a bunch of extra pixels, and you can do it with Acrobat Professional since version 8 (earlier versions will show interpolation but provide no way of turning it on). I believe Enfocus Pitstop can check for it, but cannot turn it on or off.
It is possible there are devices/readers that would honor interpolation for 8 bit data, but not 16 bit, or indexed, or ICC tagged, etc., so you would want to test heavily if this is very important.
I tend to agree with Stephen. Interpolation turns sharp pixel edges into blurry gradients, and I personally believe that the low-resolution look is more forgivable than the I-took-a-blurry-out-of-focus-image-and-slapped-it-on-my-newsletter look.
Gordo, I think you might have a point for limited resolution increases - particularly for the fancier algorithms that are almost interpolation and sharpening combined.
To change it with Acrobat, you need to create a preflight with a single "fixup." I've attached a screen shot of the fixup that gets it done.
Attachments
Trying a different method here.
Stephen Marsh
Well-known member
Gordo, this is why I am asking about expectations. Bicubic resizing of photo content in itself does not create a silk purse out of a sows ear. If the expectation is not for a miracle, then that is great.
​Yes, with a noise free and well focused digital original, one can bicubic resize up at least x2 without noticeably degrading the image too much.
For raster content that was originally vector such as logos, then fractal based resizing methods can work great and provide much better results than bicubic based methods.
I agree that output often levels things, what we see onscreen is not the same as print results. It is next to pointless pixel peeping on the monitor, even with high density Retina or similar displays that can simulate a high dpi value.
It is also very much image dependent, content of the image, how much lossy compression has been applied etc. For photographic content, there is usually little to be gained, except a slight softening of the image and potentially artefacts generated by the resampling method (which may or may not be classed as “better” depending on various factors).
Sometimes doing a x2 nearest neighbour upsample and adding a little bit of blur is similar to a bicubic resize when the original is low resolution (GIGO). EDIT: One could probably just introduce a minor blur then sharpen to the original low res image and the halftone output would look similar to the x2 bicubic upsample.
You wrote that upscaling can help to reduce or remove staircasing (aliasing). I wrote that upscaling can make the image a little softer. I believe that we are both saying the same thing in slightly different ways. Let me change “no practical gain” to “limited practical gain”.
Stephen Marsh
​Yes, with a noise free and well focused digital original, one can bicubic resize up at least x2 without noticeably degrading the image too much.
For raster content that was originally vector such as logos, then fractal based resizing methods can work great and provide much better results than bicubic based methods.
I agree that output often levels things, what we see onscreen is not the same as print results. It is next to pointless pixel peeping on the monitor, even with high density Retina or similar displays that can simulate a high dpi value.
It is also very much image dependent, content of the image, how much lossy compression has been applied etc. For photographic content, there is usually little to be gained, except a slight softening of the image and potentially artefacts generated by the resampling method (which may or may not be classed as “better” depending on various factors).
Sometimes doing a x2 nearest neighbour upsample and adding a little bit of blur is similar to a bicubic resize when the original is low resolution (GIGO). EDIT: One could probably just introduce a minor blur then sharpen to the original low res image and the halftone output would look similar to the x2 bicubic upsample.
You wrote that upscaling can help to reduce or remove staircasing (aliasing). I wrote that upscaling can make the image a little softer. I believe that we are both saying the same thing in slightly different ways. Let me change “no practical gain” to “limited practical gain”.
Stephen Marsh
Last edited:
Stephen Marsh
Well-known member
Kyle, the old forum software treated .jpg and .jpeg extensions differently allowing a large pixel size on one but only a smaller pixel dimension on the ohter and forcing a resize, as well as other formats such as .png etc. Looks like the new forum sw has different options.
Stephen Marsh
.
Stephen Marsh
.Attachments
Last edited:
David Milisock
Well-known member
My solution is a bit out of the PDF realm in that in my opinion Adobe products are the last place to do this. I perform this task daily in the large and grand format world. I take the PDF and extract the image in any way that suits, open it in Corel Photo-PAINT and export for Photozoom Pro it's a free add in in X6 and X7 and up sample the image there and save as a TIF. The end result in terms of quality depends on the quality of the original and how far you increase the size. A great image at 300 DPI can be enlarged in most cases 300% at 300 DPI. Photozoom Pro can also be bought as an application itself.
Stephen Marsh
Well-known member
David, once you have the image resized using the “secret sauce” of Photozoom Pro, then what? How do you get the image back into the PDF?
Stephen Marsh
David Milisock
Well-known member
​Ok My work flow is quite a bit different because for the most part it's large and grand format printing or 1 or 2 page press or digital press print work and I avoid Adobe products unless it is a must. It can get funky if the file has overlapping transparency that has not been flattened. Having your ducks in a row in terms of color management is a must! We only open PDF files in Acrobat using Pitstop, convert all fonts to curves rename and save the file as a working file. Depending on the results of our look at the file we many times just open it up in CorelDraw. Select the image right click edit bitmap and save the file. Unfortunately many times an image in the PDF has already been placed at the physical size and the resolution is low. In that case I resample to high res reducing the physical size properly and may print a small test to see if it's worth the effort. If the file is ok I open it in Corel Photo-PAINT and with Photozoom Pro size and res it up save as a tif use it as a replacement in the CorelDraw file and publish to PDF from CorelDraw. An all Adobe Pitstop solution is possible but with most things Adobe it's more time consuming and a PITA. As long as you know the rules and follow them it's works fine using Adobe or Corel but for 1 and 2 page files Corel is just a lot faster. The Adobe solution would to use touchup to get the image out of the PDF save the file as a tif, If you wanted a real pro up sizing then open in Corel Photo-PAINT size and res up in Photozoom Pro save as tif, open that tif in PS insert a into the pdf copy and paste the new resized image deleting the old low res. I don't work much with Adobe so there my be a better way. You could just up res it in PS but that's a POOR choice.
Last edited:
Similar threads
