Stacking error: input images have different precision

Software Siril
1

Hi, I’m using the latest Siril release and having an issue. I took a set of about 880 images previously calibrated with flats and darks but not debayered, captured across multiple days and created a new FITS images sequence, including debayering. These have the prefix pp_. Then I registered them, giving the new sequence a prefix r_. When I tried to stack, I got this error message:

12:36:20: Reading FITS: file r_pp_light_00851.fit.fz, 3 layer(s), 6016x4016 pixels, 32 bits
12:36:20: Sequence loaded: r_pp_light_ (1->881)
12:36:38: Running command: stack
12:36:38: Stacking sequence r_pp_light_
12:36:38: Processing all images in the sequence (878)
12:36:38: Stacking result will be stored as a 32-bit image
12:36:38: Computing normalization...
12:36:38: With the current memory and thread (8) limits, up to 4 thread(s) can be used for sequence normalization
12:36:38: Normalization computation time: 31.42 ms
12:36:45: Stacking error: input images have different precision
12:36:45: Opening image 544 failed
12:36:46: Stacking failed.
12:36:46: Stacking failed, please check the log to fix your issue.
12:36:46: Execution time: 7.59 s

Indeed, some of the original images I found were 16 bit and some were 32 bit (I’m not quite sure why but since they were processed on separate days I suppose the settings changed somehow), which is probably the reason for the error, and the bit format (either 16 or 32) was maintained in the debayered and registered files, e.g.,

...
09:35:28: Saving FITS: file r_pp_light_00472.fit.fz, 3 layer(s), 6016x4016 pixels, 32 bits
09:35:29: Found 636 stars in image 397, channel #1
09:35:29: Matching stars in image 397: done
09:35:29: Initial pair matches: 134
09:35:29: Pair matches after fitting: 129
09:35:29: Inliers:      0.963
09:35:29: scaleX:       1.005
09:35:29: scaleY:       1.006
09:35:29: scale:        1.006
09:35:29: rotation:  -178.375 deg
09:35:29: dx:       -6304.75 px
09:35:29: dy:       -4228.93 px
09:35:29: FWHM:        10.67 px
09:35:29: roundness:    0.79
09:35:30: Saving FITS: file r_pp_light_00549.fit.fz, 3 layer(s), 6016x4016 pixels, 16 bits
09:35:30: Reading FITS: file pp_light_00473.fit.fz, 3 layer(s), 6016x4016 pixels, 32 bits
09:35:30: Findstar: processing for channel 1...
...

Is there a way for me to convert the 16-bit files to 32 bit in batch?

Thanks,
Andrew

2

Hello

No, there isn’t. In fact, converting from 16-bit to 32-bit is just a loss of disk space.
I’d like to better understand how you came to have two different bit depths. And then recalibrate the 16-bit images so that they are 32-bit. Because normally, in Siril, if there’s a division by the flat, then the images are converted to 32-bits (if the preferences are set to 32-bits).

3

Thanks for the reply. I’m not certain how I came to have 16-bit images. I suspect I switched the setting to 16-bit unintentionally for some time (or maybe it reverted to 16-bit when I reinstalled Siril one time? not sure…) so I was producing 16-bit calibrated images until I noticed and switched it back to 32-bit. I could check the dates to see if that matches.

What do you mean it’s a loss of disk space?

I was able to write a python script using astropy to convert the 16-bit FITS files to 32-bit in batch. Happy to share if it would be helpful. The files seem to play well with Siril after conversion.

Thanks,

Andrew

4

it’s a loss of disk space because it’s the same data but taking twice the space on disk. Maybe you have processed some uncalibrated images too, in that case they stay in 16 bits