PhotometricColorCalibration background neutralization

[johnpane]

Background neutralization seems essential to obtaining a color-balanced image from PhotometricColorCalibration. And, setting an upper limit on the neutralization seems necessary to preserving color in structures that are low intensity (i.e. just above background noise). I find it very challenging to perfect this.

First, does the background neutralization limit apply before or after the photometric calibration is run? If the color scales change during calibration the user cannot easily anticipate the proper value. If it does not already do so, would it make sense for PCC to calculate the mask (or whatever information it is using to determine eligible background) prior to the photometric calibration step?

Second, prior to calibration the three channels may not have the same background levels. If setting an appropriate upper limit is critical, might it be preferable to be able to set these limits independently by channel?

Thanks,
John

 

[johnpane]

Looking at the source code, I see that the background reference is calculated after the white balance is applied. In order that the user can plug in numbers based on the image before processing, I suggest calculating the background reference prior to applying white balance to the image, applying the white balance adjustment to the calculated background reference (B0), and then applying the background neutralization after white balance is applied to the image.

Also, background neutralization subtracts more signal than is necessary to establish a neutral background. This can be avoided by subtracting the minimum of the B0 values from all three B0 values (B[0], B[1], and B[2]). I have found that running PCC without background neutralization and then doing this neutralization manually using PixelMath enables me to preserve more low-signal data and the color of that data.

The above change would also have the advantage of applying less distortion to the carefully calculated white balance of the stars.

With these changes, the suggestion in my final paragraph above may still be useful if the image is wildly out of color balance prior to running PCC.

 

[johnpane]

@Juan Conejero, I submitted merge request PCL!60 to implement these suggestions (except the idea of having separate background reference limits by channel).