We do not have IPP functions you mention which support 64-bit integer images (resize, rotate, filters). Do you need signed 64-bit integer images or unsigned?
Vladimir, signed or unsigned could be Ok . Also we could use a float but we expect better performanceusing integers and not need to go back to int after the IPP functions.
In fact we need at least 50 bits but no more than that,so anything above this size is OK. Perhaps is`possible to use a combination of 16 bits and several channels but we didn`t see much information of how to do that on the documentation. Rgds Alberto