The colour synthesis issue is rather complex and depends on many factors. The colour filter grid of the CCD can be used directly to derive primary colours by taking 4 pixel groups (yellow, cyan, magenta and green) and solving 3 or 4 simultaneous equations for red, green and blue. This simple method was used in the MX5-C software, but the MX7-C software uses a larger radius of pixels to calculate each colour value. This results in smoother colour and much less of a tendency for colour errors around stars etc. Both of these methods need to have a clearly defined filter pattern with the correct sequence of colours to permit correct decoding.
I will now describe how I construct the 'raw' file:
1) 'Fast' mode: In this mode I send two 'read' pulses (odd and even lines) to the CCD at the end of the exposure and then read out the result to an array. This gives a line sequence which sums the filter data into line pairs as per the following pattern:
line 1: Cy+G, Ye+Mg, Cy+G, Ye+Mg ...
line 2: Cy+Mg, Ye+G, Cy+Mg, Ye+G ...
Because this data represents only 290 lines in total, each line is used to create colour and luminance data and then the mean values of line pairs are inserted between the original lines to make up a 580 line image.
2) 'Interlaced' mode: As interlaced images are created from two exposures, a high resolution image is generated by reading the first exposure as in 'Fast' mode, and the second exposure similarly, but with the filter summation displaced by 1 line. This results in the second field being composed of the same filter pattern as the first, but with a 1 line vertical shift. The second field is inserted between the lines of the first field so that a 580 line frame is generated, but with 'genuine' 580 line resolution.
These images will colour decode with the same equations as 'Fast' mode, as long as the Luminance data is not generated by averaging line pairs (so as to keep full resolution).
3) 'Progressive' and 'Self guide' modes: In both Progressive and Self-guide, only one set of filters is available for each of the image fields. This is because in progressive only one image is downloaded in two readouts (odd lines and then even lines) and the same happens in self-guide as the even lines self guide for one half and the odd lines for the second half. Despite the lack of dual fields for creating an 'Interlaced' image, it is still possible to do so by adding the lines together in pairs. i.e. 1a + 1b, 1b + 2a, 2a + 2b etc. You can therefore create a 'synthesised' interlaced frame, which has the same filter structure as a true interlaced frame and will decode to colour with the same equations.
From the above, you can see that only two forms of colour decoding are needed to cover all possible forms of raw image. One for 'Fast' mode and one for 'Interlaced, Progressive and Self-guide' images. As a matter of fact, the 'Fast' mode will work for all images, but vertical resolution will be lost in the 'Interlaced' versions.
The 'Pixel' and 'Line' offsets are needed if there is any displacement of the image, relative to the nominal readout format. This is not normally needed for the VB software and the effect of using either is generally to produce an almost colourless image, with greenish speckles.