I recently tried NLP and was really impressed with the results and its easy of use!I’m therefore trying to think up a good work-flow for scanning my negatives using a film scanner and inverting with NLP.
Previously, I would scan at 10,000 DPI, save to TIFF, and downsample to 4000 DPI in Photoshop. However, since the NLP workflow suggests scanning and saving as DNG, the only option for downsampling is to use the “size reduction” option in VueScan (as post-processing DNG is not possible). The issue here is that VueScan uses a “nearest-neighbour”-type algorithm, wherein it linearly averages the color of any grid of X times X pixels. According to some old threads I’ve stumbled upon online (e.g. this), this linear downsampling produces inferior results contra downsampling via the bicubic sharper algorithm.
Has anyone found a good way of combining a better downsampling algorithm than the one found in VueScan with the DNG-NLP workflow? Are my concerns regarding size reduction in VueScan exaggarated?