Dave Beaudet
Dave Beaudet
Date: 5/12/2017 Server: 2.3 GHz Xeon x 8 cores, 24 GB RAM / Memcache 6GB Cache / Varnish 4GB malloc cache / proxy to varnish for all requests not received...
I reran the May 12 test listed above with varnish and the IIP primary cache as well as the image tile cache disabled and with 8 users as well as...
Ruven has indicated that native Intel Lanczos code is being successfully integrated into IIP and provides an additional 7-8x speedup w.r.t. the resampling operations. That processing is perhaps 40% of...
I should note that the IIP bilinear algorithm operating on the 344 to resize to 325 is virtually indistinguishable from using imagemagic with a bicubic mitchell filter (once the image...
FYI - I'm experimenting with integrating the FreeImage library into IIP in order to get access to that library's range of re-scaling options. I was also able to implement a...
I've taken Filters.h and Filters.cc from the FreeImage project and re-implemented the resizing algorithm to use scaled integers for arithmetic rather than doubles. After optimization, I'm getting impressive quality improvements...
I've done some additional performance testing and jumping to the next higher resolution with Lanczos filter really does provide the best overall results in image quality. It dips the throughput...
e.g. with regular sampling and Lanczos filter:  and with x2 oversampling and Lanczos filter: 
That's awesome Ruven! I found the FreeImage code to be the easiest to understand and rather than just using it to study how the Lanczos filter worked, I decided to...