Image Statistics

[Simon Hicks]

Hi PI Guys,
                    I have been using DSS to stack images and then process with PI. I have been looking in detail at the various dark, flat and bias frames. With regard to bias frames, I have taken a number of individual images with my Canon 400D at 1/4000th second. I have imported the CR2 files straight into PI, extracted the Luminance and then measured the mean and median of each image using the Image/Statistics module.

However, if I stack in DSS whilst using these very bias frames, then DSS creates a MasterOffset file for me in tif format. If I import this into PI and measure the mean and median the values are 2 or 3 times higher than the individual frames.

I was expecting them to be roughly the same.

I have talked to Luc at DSS about this and he cannot seem to find anything wrong with the process I am following. I am not using any white balance or anything else that would seem to adjust the images whilst combining them.

I am either doing something wrong with the way I am handling the files, or the way DSS is handling them or the way PI is giving its results. I suspect the first! :-)

Can you help?

Cheers
             Simon

[Carlos Milovic]

Hi Simon

Usually the luminance yields lower values than what we might suppose at a first sight. It is nos a simple average of the values, so I would not be much surprised if the master bias reads differents values... Make the following experiment: load the RAW files, and then calculate the statistics. Average the stats for all three channels and then compare this result with the master bias. If I'm right with my assumption, now this should be much closer.

[Simon Hicks]

Hi Carlos,
                   It seems that Luc at DSS has found out what my problem has been. Mainly it is that I don't have a PhD in image processing like all of you guys! :-)

He got me to convert all my CR2 bias offset frames to greyscale 16bit tifs, then to do the same with a light frame. Then I ran all of these through DSS. I then compared the mean and median of the resultant MasterOffset with the individual 16bit greyscale tif offsets, and they were all very close.

So it was mainly due to the different colour spaces being used and the effect this has on the mean and median.

So this also proved DSS was working correctly (as if it needed me to pprove that !!!!).

Many thanks for looking into this Carlos, your help is always appreciated.

Cheers
             Simon