“Function” parameter of the task CCDPROC

smp wrote on Apr 28, 2009

Hi,

(1)
In IRAF task CCDPROC, I set the prameter function = "mean".
And did the overscan subtraction from my raw image.

(2)
I have done this part manually to check/verify part (1):
Then I calculated mean of every horizontal line (x-dimension) within overscan region of the raw image and subtracted this mean from each pixel value of that line of this raw image.

But I found that:

Pixel value(raw image) - Mean (calculated from imstat) is not equal to pixel value (from overscan using ccdproc) of processed image

What is going wrong?

Regards
Swapnil

---

Francisco Valdes wrote on Apr 28, 2009

The choice of "mean" means to compute the average of the overscan independently for each line and subtract it from the line. This is to allow correcting data where the overscan (electronic bias) is jumping from line to line. It is not the same as the mean of the overscan strip. To get a single number for the overscan you use a fit of order 0 (which is a constant and is effectively the mean). By the way, when doing line-by-line subtraction it is better to use "minmax" than "mean" since this will eliminate occasional outliers.

Frank Valdes

---

smp wrote on Apr 28, 2009

Thanks.

I would like to know the following :

(1) If I use function parameter = "legendre", with order = 1,
independent variable is row number and dependent variable is mean of the pixel values of pixels in the row, in the overscan region.

e.g. I have 2048 data points i.e. 2048 pairs of (n,f(n))
where n = 1,2,3,....,2048
f(n) = 2412, 2413, 2420, 2409, .....etc. ( i.e. mean values, each for one row)

Now, I try to fit legendre polynomial of order 2. i.e. P2(n) = (1/2)*(3n^2 - 1)

But the graph of P2(n) looks like a curve (valley) for |n|<=1, and ranging from -0.5 to +1.0.

But the mean bias values do not change like this (decreasing upto a minimum and then increasing, a bowl type curve).

So I have not understood how CCDPROC fits legendre function to a set of mean values of the rows within overscan region.

Regards
Swapnil

---

Francisco Valdes wrote on Apr 28, 2009

Hello Swapnil,

I am still not clear what you are doing. Please post a listing of the ccdproc parameters (lpar ccdproc). What is not clear is, are you using the interactive fitting option to see the curve?

One thing that I don't understand is that in IRAF a legendre function of order 1 is a constant. To get a dip you would need an order of 3 (quadratic).

Frank

---

smp wrote on Apr 28, 2009

Sorry I was on leave, so I could not post reply.

I will reply soon, regarding your above post.

Thanks and Regards
Swapnil

---

Last post on Apr 28, 2009