Will it be more cost efficient to use DSP/FPGA, instead of x86, to add simple watermark on JPEG/GIF/... files?

My current project is to add simple logo and some sentences on the gi开发者_运维知识库ven graphic files. Each file is less than 500k on average. I am going to put the service online. So it should be able to handle 50requests per second. Our current budget is limited. Any suggestion?

Start with OS-based (x86) image modification. Test the performance. If it's not up to snuff, then you can evaluate other approaches.

While an FPGA- or DSP-based approach may be faster, it's harder to find preexisting modules or libraries. You'll end up doing more work for gains that may not be worthwhile.

You should be able to get simple watermarking implemented very quickly, so you can get some performance numbers before all your functionality is present.