This particular case study is an experimental trial

Transcription

1 C38 (Needle Galaxy) You live and learn. At any rate, you live. Douglas Adams First Light Assignments 213 Equipment: Refractor, 132 mm aperture, 916 mm focal length Reducer / flattener (0.8x) Starlight Xpress Trius SX694 (Sony ICX694AL sensor) Starlight Xpress Mini Filter Wheel (Baader filters) Starlight Xpress Lodestar off-axis guider 10Micron GM1000 HPS mount Software: (Windows 7) Maxim DL 5.24, ASCOM drivers Model Maker (Per Frejvall) for acquiring tracking model PixInsight (Mac OSX) Exposure: (RGB) RGB bin 1; 20 x 300 seconds each (no luminance) This particular case study is an experimental trial of the alternative and growing trend to use unbinned RGB exposures (without a set of luminance exposures) rather than the conventional LRGB sequence in which the RGB exposures are binned 2x2. Those that do, often use this approach for imaging clusters and dense starfields. The narrow filters bandwidths are less prone to saturate and can be individually focused, whereas chromatic aberration will cause the wide-band luminance signal to smear. This should enable tightlyfocused stars with good color. Since the red and green filters exclude the dominant sodium light wavelength, the shot noise arising from light pollution should be reduced at the same time. This example evaluates the technique with the bright edge-on galaxy C38 (NGC 4565) and its three companions. This galaxy has a high surface brightness and scientists believe it is has a very similar structure to our own galaxy. Given that it is more luminous than the giant M31 Andromeda Galaxy, it is surprising that it was missed by Charles Messier. Acquisition The image acquisition took place during my first light with a 10Micron GM1000HPS mount. Without a polar scope, this mount requires a multiple point model, using plate solve synchronizations, to provide the adjustment parameters for the Alt/Az bolts. I had some teething problems with autoguiding and dither settings and it took several nights to complete 3 hours of good data, as many exposures had to be rejected due to elongated stars. The William Optics field-flattener was screw-coupled directly to the back of the refractor focuser drawtube to minimize flexure. The tripod spikes rested on concrete blocks and everything was double-checked for rigidity and play. The sensor in the imaging camera has an extremely low thermal current and excellent read noise. Calibration bias, dark and flat frames were acquired during the many cloudy nights during the preceding month. Image Calibration Using the convenient batch pre-processing script in PixInsight, the individual bias and dark frames were integrated and the flat frames calibrated and integrated to form three master files. These master files were then used to calibrate the light frames. The light frame with the smallest FWHM was selected as the master reference and the script registered the other 3 sets of light frames to it, including any that were inverted after the meridian flip. The final stacking was done with the ImageIntegration tool, using the settings in fig.1.

2 214 The Astrophotography Manual R,G & B (calibrated, registered and cropped) synthetic Luminance automatc extractor noise weighted integrated RGB combine RGB PSF extraction & deconvolve inverted blurred stretched image as mask (mask 1) neutralization ATWT to reduce noise color calibration Histogram Transformation clipped mask 1 MMT ACDNR fig.1 This shows the final integration settings for the calibrated and registered red, green and blue channels. It weights the contributions according to the exposure noise level. Histogram Transformation (RangeMask) SCNR to remove green pixels Local Histogram Equalization on galaxy blurred image as mask (inverted mask 1) extract lum linear fit to synthetic lum combine scaled lum into RGB using Lab mode increase color saturation LRGBcombination (boost saturation) processed synthetic luminance MMT enhance small scale (inverted mask 1) ACDNR fig.2 This shows the RGB integration settings for the red, green and blue stacks to form a synthetic luminance. It adds and scales the contributions according to the exposure noise level but, unlike the general integration settings, does not reject pixels. fig.3 The PixInsight processing sequence for this image follows the dual path approach used before one that optimizes luminance structures and noise, and the other for color saturation and definition in light structures and in the. With only an hour s data per channel, this image required extra at key points to control the appearance. Masks were used extensively, both in inverted and non-inverted forms and with and without mild blurring, to support the multiscale processing tools for enhancement and.

3 First Light Assignments 215 Image Processing Effective image processing handles the luminance and color data separately before combining. In this case, the luminance has to be created from the RGB data. The processing sequence broadly follows the standard route with a few exceptions to suppress noise, generate the synthetic luminance and enhance the galaxy structures (fig.3). Since this image has plenty of empty sky, without nebulosity, it is a perfect candidate for the AutomaticBackgroundExtraction tool on each stack. The resulting three FITS files were stored as a key intermediate step. Synthetic Luminance Processing Luminance data resides in each of the RGB channels and it is tempting to use the cleaner Red channel as a surrogate luminance. This, though, would favor that color in the final color image and a combination of all channel data is required to generate the luminance signal. The signal levels and hence the signal to noise ratio differ quite significantly between the channels and the current wisdom averages and weights the R, G and B data according to their noise levels. This compromise may distort the image color slightly but optimizes the use of the noisy blue and green channels. Since the channels have already had Sigma clipping applied during their own integration process, further rejection is not required. The ImageIntegration parameters are subtly different (fig.2). Luminance processing now follows a more familiar path, using an extracted PSF image to deconvolve (sharpen) the image (fig.4) and a series of s with mask support to reduce the obvious noise. Noise reduction in this case was performed before and after stretching, using the ATWT (fig.5) and ACDNR tools (fig.6), both of which are multi-scale tools that target image structures of a particular size for or emphasis. After the HistogramTransformation the galaxy was quite bright but relatively featureless. This was remedied by the LocalHistogramEqualization tool through a blurred mask. The greatly improved the dust lane appearance (fig.8). fig.4 The settings for deconvolution can be convoluted! After considerable experimentation on a preview, the settings above significantly improved star definition. A mask was used to protect the as the deconvolution process can misinterpret noise and curdle the sky. fig.5 The ATWT is an effective tool when it is used on the linear image. (In PixInsight, a green underline on the view title indicates that a screen stretch is being applied.) fig.6 Once the image is stretched, the ACDNR tool works on the non-linear image in conjunction with a mask to protect light areas. In this case, it is removing some of the blotchy areas. The key parameters for experimentation are the structure size and protection.

4 216 The Astrophotography Manual Linear RGB Processing Since the channels had already had their gradients removed, the ChannelCombination tool was used to form an RGB image, followed by the Background- Neutralization tool to remove the orange color cast. The image was still distinctly warm and a preview placed over the galaxy was used to set a neutral color reference for the entire image for the ColorCalibration tool. The MMT tool, set up for on small structure scales and through an inverted luminance mask reduced the color noise prior to image stretching. Non-Linear RGB Processing The RGB image was stretched in two passes to manage noise. The first pass used a mask made from a softened version of a stretched inverted clone image. This increased the brightness of the stars and galaxy and protected the. A second less severe stretch without the mask set an accurate shadow clipping point and the overall correct appearance. Even so, there were many green pixels in the and these were quickly removed with the SCNR tool. At this point, although the RGB and Luminance images looked similar, they required balancing before doing the LRGB combination. This was accomplished using the LinearFit tool. First, the luminance information was extracted from the RGB file. The LinearFit tool was applied to this extract, using the synthetic luminance file as the reference. With the two luminance files matched, the adjusted extracted luminance file was recombined into fig.8 The LocalHistogramEqualization tool is very adept at emphasizing contrast at a local scale. Using a mask, it brings clarity to the galaxy and stops it being a white blob. the RGB file, using the ChannelCombination tool set to CIE L*a*b*. Since the LRGBCombination tool process often lowers color saturation, the RGB saturation was boosted beforehand using the ColorSaturation tool. Even so, the LRGBCombination settings still required a small increase in saturation (fig.7). Since the origin of the luminance data was the RGB data itself, the lightness setting was fully applied. Final Tuning The image still required some work. The dust lane was nicely colored but lacked clarity and the noise was still obvious. In both cases I continued with multiscale processing tools to suppress and enhance structures. In the first instance, the fig.7 Applying the synthetic luminance to the RGB file requires experimentation with the lightness and saturation sliders. In this case the saturation was boosted (moving the slider to the left) and the lightness was fully applied to replace the lightness in the RGB image. fig.9 The MMT tool is set up for enhancing galaxy structures in this case. (A mask was used when applied to the main image.)

5 First Light Assignments 217 MultiscaleMedianTranform tool was applied again, only this time set up to emphasize structures at 1, 2 and 4 pixel scales. The was protected with a soft edged mask and cleaned up further using the ACDNR tool and the same mask, only this time, inverted. Conclusions I set out to evaluate unbinned RGB exposures without Luminance support. With an hour of data per channel and after image stretching to bring the galaxy to life, there was obvious noise, even with a low noise sensor. This required several additional process steps to control its level during both linear and non-linear processing. Fortunately this image had no faint nebulosity and simple masks were adequate to isolate the galaxy and stars from the. This is an interesting experiment rather than an exhaustive test, using a modest exposure. I certainly feel this approach will be very useful for dense starfields and clusters. In both these cases the required image stretch is less than that required for a dim galaxy or nebula and will equally have less impact on noise. These subjects demand small colorful stars without excessive bloat or clipping. The same approach applies for a starfield overlay to support a narrowband image. (From now on, this will be my preferred approach for adding in colored stars into an otherwise narrowband image.) I am not so sure that this technique is better than LRGB for imaging galaxies and nebula. To evaluate the true merits of L(1) RGB(2) versus RGB(1) requires a calibrated back to back test, with the same overall imaging time in each case and applied to a field of view with plenty of object variety. fig.10 The full image. There are three neighboring galaxies that are believed to be interacting with the main galaxy, labelled NGC 4565 a, b and c. This final image has good star color and shape. A few magnitude 7 stars are starting to bloat and the, after multiple s, is just satisfactory. As always, more exposure would improve things.