I'm coming over to Affinity Photo from Photoshop and hit a problem which I had to resolve quickly - that meant reverting to Photoshop because I know how to do it.
I regularly require reduced copies of photographs for websites, and with Photoshop I do this with Save For Web and Devices, using Bicubic Sharper as the resampling method. This option isn't available in Affinity Photo, which offers Nearest Neighbour, Bicubic, Bilinear, Lanczos 3 (separable) and Lanczos 3 (non-separable). I was in a hurry to do one in Affinity Photo, so accepted the default Nearest Neighbour, which produced a dreadful result, full of artefacts even saving as a JPEG 100% quality. I don't normally save in JPEG format (I prefer PNG) but it was a quick fix to replace a file uploaded from my iPad using a crappy (free) resize app which does only JPEG. It was quicker to open the photograph in Photoshop and do it there than to experiment with different settings or even consult the help system, so that job is done.
Coming back to the problem, I find that here is zero information in the Affinity Photo help system (that I could find) explaining the different resampling methods and which is best for what kind of resizing. Wikipedia is not a great help, the article on Lanczos resampling being mathematical rather than practical. I have a degree in Physics and the maths isn't a problem, but it's no help in deciding on the most effective method for image reduction for the web. It appears that the Lanczos method introduces ringing artefacts which may give the illusion of sharper edges than bicubic, so one of these methods may be similar to Photoshop's 'bicubic sharper', which I find acceptable for reductions down to 300-400 pixels wide.
Can someone explain in simple language which of the methods is most appropriate for my needs? I have a feeling that it is probably one of the Lanczos 3 methods or bicubic but it would be nice to know the ins and outs of each method.