From: Bart <sca...@gm...> - 2017-12-22 15:25:42
|
> > For 4-bit, I'd say read into uint8 and do some bit massaging (AND and shift). (ideally do it while loading, so that you don't need to effectively allocate the whole thing two/three times) And in case of UCSF, the gain reference is a NX*NY float32 image right after the agard-style extended header. You can detect its presence by seeing that NSYMBT is exactly that much (NX*NY*4) larger than NINT and NREAL would suggest. ...as far as I understand it, anyway. I've been writing a module that reads extended data for all variants that can come out of our setup, meaning FEI, UCSF, 4-bit, and SerialEM. While almost no variant marks itself, you can get suprisingly far in detection. ...though I keep on coming across new and exciting real-world deviations, so this needs more testing before it's useful to anyone else. --Bart Alewijnse @NeCEN |