Summary

This is IC 1805, the Heart Nebula (Cas) in SHO.

For this image I was interested in creating first a minimally processed image without any direct denoising, deblurring or sharpening steps. I then wished to examine the effects of denoising this image followed by an additional sharpening step.

Two data sets, taken with the same camera and telescope, but separated in time by 13 months, were combined for processing. This provided enough total integration time for the minimally processed image to have a decently high SNR in all three narrowband channels. Processing was done in PixInsight.

Minimally Processed Image

Frames for each channel were combined using a weighting function dependent on both SNR and FWHM with Winsorized Sigma clipping pixel rejection. Background gradient was removed using DynamicBackgroundExtraction with a mostly linear background model. Channels were combined simply a R=S, G=H, B=O. ChannelCalibration was used to re-balance the color channels while linear. After stretching, a series of several CurvesTransformation processes with various ColorMask produced masks were done to adjust the colors to taste as well as to adjust the background vs. nebula intensities and color saturation. This is the extent of the processing done on the Original version and the left-most images in the two mosaics.

Noise Reduction Image

The minimally processed image was then masked to protect the brighter nebula regions. TGVDN (TGV Denoise) was used for noise reduction in the Lab channels using 75% of the default settings, 100% of which were deemed to be too aggressive. This is Revision B.

Sharpened Image

The noise reduced image was then sharpened using MultiscaleLinearTransform (MLT). The image was masked to protect background, dimmer nebula regions as well as most of the stars. I did not wish for the stars to be sharpened as this effect is displeasing to me. MLT was used with 4 default layers, Bias adjustment of the layers and no Noise Reduction settings. This is a step that can easily be overblown, and for this step I chose small values spread over all four layers. This is Revision C.

Cropped Images

The Revision D mosaic shows (top row) the vicinity around Melotte 15 (center of nebula) to the left (East) edge of IC 1805, which has interesting dark nebula structure with bright background illumination. Also shown (bottom row) is IC 1795, the Fish Head Nebula, upper right region of IC 1805. Leftmost are the minimally processed versions from Original, middle with denoising from Rev B, right with additional sharpening from Rev C. Melotte `15 in pariticular shows the most effect from denoising of the surrounding area and the sharpening. I like the result, however I do note the introduction of some noise back into the image.

The Revision E mosaic is a close cropped area around a small planetary nebula called WeBo 1 (PN G135.6+01.0). I won't take time to describe it here, but those interested may wish to follow the link to read up on it. In this image it appears dominantly blue in the OIII channel. It is located left middle from the edge of IC1805. To my eye, the denoising step has mild improvement as this is a small scale object within an otherwise much wider scale image with low resolution. In this case, any sharpening has been minimal as this PN was mostly masked out from that process.

Remarks

This was an interesting exercise, and I hope you find it illuminating as well. My remarks are as follows. The minimally processed image shows an inherent uniformity, and the noise and granularity is, for me anyway, pleasantly reduced due to the larger amount of integration time from the two combined data sets. As a finished image, I think it can stand on its own.

With noise reduction (NR), background granularity is reduced. I purposely chose an amount of NR that left a small amount of noise behind as to not produce an overly processed "creamy" and unnatural look. How well I achieved this is in the eye of the beholder.

Sharpening is the step that can easily be overdone, and as I stated above, I have adopted the process of masking out stars as I don't like the results of their sharpening at all. They are often saturated enough in comparison with the nebula signal, and even a small sharpening step can cause the stars to become flat, bright discs devoid of any intensity spread. A consequence of sharpening is introduction of noise once again, and I selected MLT values such that this, at least through visual inspection, was balanced against the amount of sharpening I wished to achieve.

The amount of denoising, deblurring and noise reduction is, of course, a subjective decision, however I have determined that knowledge of both the negative and positive effects of each processing step is vital to these choices.

I hope you find the images pleasing as well as these notes.