WCMY
as of 4/14/99
Richard Berry, Al Kelly, Ed Grafton and I have been working together on developing a new technique for color CCD imaging. Instead of using the conventional Red, Green and Blue additive filters, we have worked out the ability to use Cyan, Magenta, and Yellow Subtractive filters. Richard has made significant improvemnts in his QColor.exe software package to include this processing capability as well as a number of other processing enhancements and will be releasing the software in the near future when we have completed the development work.
The details of how to use this technique are available in the December, 1998 issue of Sky & Telescope Magazine..
The really exciting thing about this is it is based on using the very inexpensive CMY dichroic filter set from Edmund Scientific. These filters not only allow a major reduction in exposure times for the color layers from what is needed with RGB filters, but also have much better coverage in critical wavelengths (like OIII, etc.) to better image emission objects.
On November 6th and 7th, 1998, Al Kelly and I had the pleasure of participating in the "Image-The-Sky-98" Conference in Salem, Oregon where we were able to further introduce this technique to other fellow ccd'ers.
I have started writing a summary of the different methods for calculating the "weights" used in the software for combining the filtered color images into a color composite to achive proper color balance. In addition, I am documenting a "step by step" type of writeup for capturing and processing WCMY (and WRGB) images using the techniques we presented in the S&T article. As I update the process, I will keep updating this website page. What I have so far follows:
So far (as of 8/23/98), Richard Berry, Chuck Shaw, Al Kelly, and Ed Grafton have identified and tested 5 different methods (and some options for each method) for compositing filtered CCD images into a color CCD image. All the testing has also used a quad color process. Namely, the filtered images are used to composite the chrominance layer and an unfiltered (or with only an IR blocker in place) White image for the luminance layer. The filtered images can also be combined without the White image to build a color image. In addition, the filtered images can be used to build a pseudo White image that can be used in lieu of a separate white image, or can be averaged with the White image to potentially improve the White image’s Signal to Noise Ratio (SNR).
The actual compositing activity has been done using development version of Richard Berry’s QuadColor software (see Richard’s website for availability). Since Ed Grafton uses an ST-6 camera (Al, Chuck, and Richard use CB245 cameras), Keith Kelly developed a program to convert ST-6 images into CB245 images so Ed’s images could be processed via QColor.
The conversion process in Qcolor to convert a WCMY filtered image set to a "synthetic" RGB set is referred to as "CO" in the following method descriptions. The function in Qcolor to combine the WRGB filtered images (or the synthetic RGB images produced by the CO function from WCMY images) is referred to as the "AC" function in the following method descriptions.
The AC function has two capabilities, in addition to its main function of compositing the RGB images into a color image. It can apply a multiplier to each pixel to apply the extinction correction (referred to as "CON" in the AC function), and it can add/subtract values to each pixel to apply the correction for the sky background (referred to as "ADD" in the AC function). ADD/CON are also available for the White image, as is Gamma (overall brightness) and Saturation (for overall color intensity) for the compositing activity.
The initial "AC" generated image from whichever method is used must be further processed by at least two more steps to obtain correct color balance. The first is to correct for "extinction" of each color as a function of its distance from the zenith. The second is to correct for light pollution/sky background effects. Both of these corrections are discussed in detail on Richard’s website and are performed by additional "AC" passes in Qcolor.
So far, Method 5A has shown the best overall results for the least efforts. A brief description of and comments about the five methods identified so far to calculate weights to be used in Qcolor are as follows:
_______________________________________________________________________________________
Method 1: This method is for use when using only RGB filters and compositing color images only from WRGB filtered image sets. The overall approach is to measure fluxes passed by RGB filters and IR blocker as detected by CCD chip, and normalize them to one of the filter’s outputs. Three options available:
Method 1A:
Measure actual RGB flux values as passed by actual RGB filters and actual IR Blocker and detected by actual CCD chip of the same "white" target (G2v star) in each RGB image, using the photometry function in Q-Color (ED function, "S" key), then correct for extinction to determine the zenith flux value, and normalize these values to Red value. Load RGB images in Q-Color via LR and use the normalized flux values in the AC function in Q-Color
Method 1B:
Load RGB images via LR. Allow Q-color to calculate the weights by using the ER function (this stretches each RGB image loaded so that the brightest object in each image is assumed to have been white, and the darkest object is assumed to have been black). Use all 1’s in the AC function.
Method 1C:
Use "ideal" flux values for RGB fluxes instead of actual measured flux values. This requires transmission curves for an "ideal" set of RGB and CMY filters to be multiplied by the transmission curve of an "ideal" IR blocker (another sub option here is to use the actual IR blocker transmission curve), and then multiply by the QE curve for the chip being used. Normalize these values to Red value. Load RGB images in Q-Color via LR and use the normalized flux values in the AC function in Q-Color
Method 2: This method is for use when you have both RGB and CMY filter sets available for the calibration activity, and are compositing color images from WCMY filtered image sets. The overall approach is to measure White (G2v star) fluxes passed by RGB and CMY filters and IR blocker as detected by CCD chip. Then combine these fluxes per the following equations to derive weights to be used in the Q-Color CO function.
Cwt = (G+B)/C
Mwt= (R+B)/M
Ywt = (R+G)/Y
Rwt = W/R
Gwt = W/G
Bwt = W/B
There are two options to this method:
Method 2A:
Use actual fluxes measured as passed by actual RGB and CMY filters, actual IR blocker, and actual CCD chip. Correct for extinction to get the zenith weights.
Method 2B:
Use "ideal" flux values for RGB and CMY fluxes instead of actual measured flux values. This requires transmission curves for an "ideal" set of RGB and CMY filters to be multiplied by the transmission curve of an "ideal" IR blocker (another sub option here is to use the actual IR blocker transmission curve), and then multiply by the QE curve for the chip being used.
Method 3: This method is for use with only a CMY filter set (i.e. no RGB filters available). The overall approach is to measure fluxes passed by actual CMY filters and actual IR blocker as detected by actual CCD chip, and normalize them to one of the filter’s outputs.
There are three two options available:
Method 3A:
Measure actual CMY flux values as passed by actual CMY filters and actual IR Blocker and detected by actual CCD chip of the same "white" target (G2v star) in each CMY image, using Photometry function in Q-Color (ED function, "S" key) and normalize these values to Yellow value. Load CMY images in Q-Color via LC and use the normalized flux values in the CO function in Q-Color for the CMY weights. Use 1’s for RGB weights. Use default values in AC. Note: one of the major problems with this method is that there is no clean, direct way to correct for RGB extinction to get zenith values.
Method 3B:
Load the CMY images into Q-color via LC. Allow Q-color to calculate the weights by using the EC function (this stretches each CMY image loaded so that the brightest object in each image is assumed to have been white, and the darkest object is assumed to have been black). Use all 1’s in the CO function and all default values in the AC function.
Method 3C:
Use "ideal" flux values for CMY fluxes instead of actual measured flux values. This requires transmission curves for an "ideal" set of CMY filters to be multiplied by the transmission curve of an "ideal" IR blocker (another sub option here is to use the actual IR blocker transmission curve), and then multiply by the QE curve for the chip being used. Normalize these values to Yellow value. Load CMY images in Q-Color via LC and use the normalized flux values in the CO function in Q-Color for the CMY weights. Use 1’s for RGB weights. Use default values in AC.
Note: For the cases of "ideal" filters, there is no distinction between Methods 5C, 4B, and 3C….the weights would be the same. This is why there aren’t REALLY that many options, but they are listed as separate options for the sake of completeness.
Method 4: This method is for use when an RGB filter set is available for the calibration activity and the images are to be composited from CMY filtered image sets (Method 4A), or when only CMY filters are available (Method 4B). The overall approach is to use fluxes passed by RGB filters (real or "ideal") and IR blocker as detected by CCD chip in the following equations to calculate the CMY weights to be used in Q-Color CO function. RGB weights should be all 1’s in the CO function. Use all default values in the AC function.
C = B+G
M = B+R
Y = G+R
Cwt = C/Y
Mwt = M/Y
Ywt = Y/Y
There are two sub methods available:
Method 4A:
Use flux values from G2v star as passed by actual RGB filters and actual IR blocker as detected by actual CCD chip.
Method 4B:
Use "ideal" flux values for RGB fluxes instead of actual measured flux values. This requires transmission curves for an "ideal" set of RGB filters to be multiplied by the transmission curve of an "ideal" IR blocker (another sub option here is to use the actual IR blocker transmission curve), and then multiply by the QE curve for the chip being used.
Note: For the cases of "ideal" filters, there is no distinction between Methods 5C, 4B, and 3C….the weights would be the same. This is why there aren’t REALLY that many options, but they are listed as separate options for the sake of completeness.
Method 5: This method is for use when both the RGB and CMY filter sets are available for the calibration activity and the color image is to be composited from WCMY filtered image sets (Method 5A) or when only the CMY filter set is available (Methods 5B and 5C). The overall approach is to use fluxes passed by RGB (real or "ideal") and CMY filters and IR blocker as detected by CCD chip in the following equations to calculate the CMY weights to be used in Q-Color CO function. RGB weights should be all 1’s in the CO function. Use all default values in the AC function.
Cwt = (B + G + C)/Y
Mwt = (B + R + M)/Y
Ywt = (R + G + Y)/Y
There are three options to this method:
Method 5A:
Use flux values from G2v star as passed by actual RGB and CMY filters and actual IR blocker as detected by actual CCD chip.
Method 5B:
Use "ideal" flux values for RGB fluxes instead of actual measured flux values. This requires transmission curves for an "ideal" set of RGB filters to be multiplied by the transmission curve of an "ideal" IR blocker (another sub option here is to use the actual IR blocker transmission curve), and then multiply by the QE curve for the chip being used. The actual fluxes as passed by the actual CMY filters and actual IR blocker as detected by the actual CCD chip should still be used.
Method 5C:
Use "ideal" flux values for RGB and CMY fluxes instead of actual measured flux values. This requires transmission curves for an "ideal" set of RGB and CMY filters to be multiplied by the transmission curve of an "ideal" IR blocker (another sub option here is to use the actual IR blocker transmission curve), and then multiply by the QE curve for the chip being used.
Note: For the cases of "ideal" filters, there is no distinction between Methods 5C, 4B, and 3C….the weights would be the same. This is why there aren’t REALLY that many options, but they are listed as separate options for the sake of completeness.
Al Kelly and Ed Grafton and Richard Berry and I have posted several of the early images that have been produced using this technique and I invite you to visit our websites to see the results so far......
Keep Tuned!!!
(Please use your Browser's BACK button to return to the CCD page)