›

Atik 16200 column defects

Ed Stuckey
in General
Adam,
I know how to remove column defects. Darks and Bias frames are self explanatory. Do I need to remove the defects in flat frames and light frames, or are these averaged out during calibration.
Thanks,
Ed Stuckey

Comments

  • BlockHeadBlockHead
    Column defects probably should be taken out after the files are calibrated. Since the defect exists in the darks and flats- giving the calibration process the chance to do its thing first- then if those pixels are still no good- do the defect repair. So do this only on the calibrated light frames. If defect removal (either CC or defect map) doesn't entirely take care of the issue... dithered data will... *and* if you have dithered data *and* statistical rejection doesn't take care of it.. you can do selective rejection as a last resort!
    -the Blockhead
  • Ed Stuckey
    Thank you, excellent explanation.
  • BlockHeadBlockHead
    One more thing... 
    There is also a script call Linear Pattern Subtraction... it can be useful too.
    I will be highlighting this script soon. If you have many column defects..this might be the answer.
    -the Blockhead
  • Andrew J
    Hello Ed. 

    I also have a ATIK 16200 with several bad columns on the sensor that image integration would not remove even when using a high dither. Unfortunately, the Defect Map method outlined in Adam's video would not effectively remove these columns. I had to use the Linear Defect Detection (LDD) and Linear Defect Subtraction (LDS) scripts to remove the columns from my calibrated frames. The LDD script is used first to create a "map" of the bad columns by using image integration to combine a set of Unregistered calibrated frames. This will create an image where all the bad columns are aligned and give enough signal for the LDD script to work effectively. The good news is that you should only have to run the LDD script once (or when additional bad columns show up on your sensor). 

    Note, You maybe tempted to combine the line items in the LDD Bad Column map file as you may notice the same column appears more than once but with different ranges of Y pixels. Don't. Apparently there is a reason that the LDD script breaks up some columns. When I manually combined them in the bad column map text file the LDS script would not work effectively. I believe this is because the LDD script detects a different amount of correction required for different "sections" of the bad column. By leaving the bad column map as generated by the LDD script seems to work best.

    With my sensor, even the LDD/LDS scripts by themselves was not enough to correct all the bad columns in my data. It was only after I reached out to Gerald Wechselberger, one of the people who helped create the LDD and LDS scripts, did I finally get a solution. For 1 column on my sensor, the only way I can get the LDS script to correct the bad column is by applying the pixel math formula below to each calibrated light frame using an image container before running the LDS script. This formula simply writes "0" into the portion of column 631 that was not being corrected. Once I did this then the LDS script works almost perfectly and unless I am really looking I cannot detect any bad columns in my data. If Gerald ever reads this, thank you again for your help!!! 

    The Pixel Math formula: 

    // RGB/K:

    d = d2seg( x1, y1, x2, y2 ); iif( d < r, 0, $T )


    // Symbols:

    x1 = 631, y1 = 2275, x2 = 631, y2 = 3596, d, r = 1 

    Finally, when I run the LDS Script on my calibrated frames with the corrected column, I do not use "Correct the entire image". When I used this feature it seems to over correct and leaves artifacts. I found that just having the LDS script work on the columns defined in the defects file created by LDD seems to work best. YMMV. 

    The settings for LDS that have worked best for me are 
    Layers to Remove: 9
    Rejection Limit: 3
    Global Rejection: checked
    Global Rejection Limit: 7
    Background reference region: A region in the images that has representative background sky for the entire column. Mine is 680 x 50 x 36 x 3498.

    Hope this helps.

    Andrew J
  • Ed Stuckey
    Thanks, Andrew. I have not been on the forum recently, so just saw you post today. I will try your method.
    Ed
  • Ed Stuckey
    one more question, what if you might have a bad row. I cannot see it in an individual say dark frame, but when I calibrate, register, and image integration it shows up as a dark streak running horizontally across part of the frame. I know this is sign of a problem because I ran column defect on a really obvious column defect and it removed the vertical dark streak in the integrated image. I have one more vertical streak caused by a defective column that column defect would not remove. I am going to try andrew’s method. 
    What is believe I observed is if the defective column is really prominent then column defect will resolve the issue,however if the defective column is barely visible the column defect removal does not work.
Sign In or Register to comment.