MRC files with signed bytes are not read correctly

Tags: #<Tag:0x00007fd547f37ef0> #<Tag:0x00007fd547f37c48> #<Tag:0x00007fd547f37978>


I am the principal developer of the IMOD package, which uses MRC format. It now writes byte files as signed bytes by default, following a clarification of the MRC standard about 5 years that was formalized in a recent publication (Cheng et al. (2015) MRC2014: Extensions to the MRC format header for electron cryo-microscopy and tomography. J. Struct. Biol. 192, 146-150.). Users have recently reported that signed byte files do not read correctly (are not recognized as signed) when read into ImageJ through the scifio2 library or the regular import mechanism by drag and drop. They do read correctly when read in with the Bio-formats plugin.

It would be great if the files read correctly through whatever mechanism is used.


@David_Mastronarde Thanks for taking the time to report this here!

Do you have a sample non-working file you could share?

is an example of the same data signed and unsigned. (Your system wouldn’t let me upload). The signed one happens to have the MRC standard field set at 10240 at offset 108 (154 octal) but older files and other signed files from IMOD with non-zero origins currently will not have this set.


Thanks @David_Mastronarde. I will take a look when I have time—probably in late July or August. (Sorry for the long delay, but it is an extremely busy time right now.)