big changes in intensity range after bilinear interpolation in non-rigid registration

[pwang724]

After non-rigid registration I noticed that the pixel intensity ranges are dramatically different compared to the raw image. I took the raw fluorescence by multiplying the binary roi mask (unweighted) with either the raw tifs or with registered tifs outputted by suite2p after non-rigid registration (reg_tif = True). I found that the fluorescence measurements significantly differed between the two. My expectation is that intensity values should be more or less halved, as per tiff_to_binary.py, but they are instead higher. For example,

This is the fluorescence from raw images:

This is the fluorescence from non-rigid registered tifs:

Moreover, a significant amount of pixel intensities are negative.

Is this some odd behavior from bilinear interpolation? Is this reproducible in your hands?

[carsen-stringer]

could this be when the tiff is saved it is a different format? have you checked the fluorescence values from stat?