This is a great question. I just spent some time testing a high-res image (600ppi) with fine textures and sharp edges. I resampled it to 72ppi both from within Ps and from within ID (when exporting to PDF). I turned off Compression just to reduce the variables.
First, the differences between the top three options as I explain below was very small, especially for such a large resampling job (600 to 72). So I wouldn't bother doing resampling in PS unless a) the photograph was the most important part of a story, like a coffee table book with huge close-ups of antiques, for example; or b) the photograph was changing its ppi by more than 50% or so. For the majority of images (like 95%) InDesign's own algorithms when exporting to a press-ready PDF will be fine. For ones that'd be on the web (e.g. PDFs of a guide or newsletter or whatever for downloading from a web site), there's virtually no need to jump back into Photoshop to resample.
- What I found is that the best resampling was done by PS using Bicubic Reduction (Sharper Edges).
- Second best was a tie between PS using the "default" Bicubic (which says "smoother gradients" but I've never seen evidence of that) and ID using Bicubic. These might be exactly the same. A very small amount of detail in the fine textures was smoothed over in the ID example.
- Third best was ID using Average Downsampling, same amount of details were smoothed, and some sharp curves showed some tiny jaggies when you zoomed in.
- Way way down in fourth place was Ps using Nearest Neighbor, which I did just as a test. ;-)
I have a PDF showing the results, lmk if you want a copy.
Here is what Adobe's documentation says about InDesign's resampling options when exporting to PDF:
Average Downsampling To Averages the pixels in a sample area and replaces the entire area with the average pixel color at the specified resolution. Subsampling To Chooses a pixel in the center of the sample area and replaces the entire area with that pixel color. Subsampling significantly reduces the conversion time compared with downsampling but results in images that are less smooth and continuous. Bicubic Downsampling To Uses a weighted average to determine pixel color, which usually yields better results than the simple averaging method of downsampling. Bicubic is the slowest but most precise method, resulting in the smoothest tonal gradations. |
AM