The screenshot below shows the ilastik display of a 16-bit TIFF on the left and the correct display (using vigra and VigraQt) on the right. 16-bit PNG is even worse -- I just got a white image. The problem is the same in both OpenGL and software rendering. Looking at the colors values displayed in the status bar gives some information about what might happen:
in the TIFF case, the file contains unsigned short, but it seems that data is read as if it were signed short and than normalized (i.e. shifted) to the range 0...65535. The log in the console window says Normalizing your data for display purpos, I assume you have 12bit data, which is, of course, wrong.
in the PNG case, the file also contains unsigned short, which seems to be read correctly (according to the status bar values), but is not properly normalized for display. Interestingly, the preview thumbnail in the "Project >> Add" dialog is correct.
The test images I used: TIFFPNG (sorry, these files are huge).
The screenshot below shows the ilastik display of a 16-bit TIFF on the left and the correct display (using vigra and VigraQt) on the right. 16-bit PNG is even worse -- I just got a white image. The problem is the same in both OpenGL and software rendering. Looking at the colors values displayed in the status bar gives some information about what might happen:
Normalizing your data for display purpos, I assume you have 12bit data
, which is, of course, wrong.The test images I used: TIFF PNG (sorry, these files are huge).