15
« on: July 25, 2010, 06:45:08 PM »
I see that the manual states that one should, "se the original files from your camera whenever possible". Is it acceptable to downsample images? When working with Photosynth or Bundler, I regularly downsample images to 2 or 4 megapixels respectively. On my computer, SIFT does not process images larger than 4 megapixels. I see that PhotoScan will align the 8 megapixel images that are produced by my Canon S90 camera. I found that PhotoScan will in some cases also process images that are downsampled by the ImageMagik command "mogrify -resize". However, in other cases downsampled images do not appear to solve well. Is there a preferred method of downsampling, or is there a preferred ratio--say 50%.
I suspect that downsampling images permits the processing of larger collections. At a full 8 megapixel size, there is more information in the photos than is required. Downsampling, when it works, results in faster processing and models with acceptable levels of detail. I would be very interested to learn if there are any suggested "best practices" when it comes to software used and downsampling factor applied. Any help would be appreciated.
Regards,
Nathan