do you think about how it is leaving the matter? I need a perspective back outside to see if I'm on the right track, so you can leave your comments here.
The following link is the starting image:
http://pteam.pixinsight.com/sonnenstein/sagitario_raw.jpg
As I mentioned, the processing of this image was made using a criterion rather than the other images that I processed these days.
And in this other link shows the result after processing:
http://pteam.pixinsight.com/sonnenstein/sagitario_test.jpg
The first step, first of all consists of a single halftone setting really aggressive in the channel RGB / K combined (0.00001). This setting delivers pixel values \u200b\u200bin the entire image around 0.9995!
then adjust the dynamic range ( autoclip in the shadows and lights) the sky background starts to have values \u200b\u200baround 0.4 to 0.6, as shown in the following video:
This first adjustment is not always the work on CCD images, because the raw data and tend to occupy the entire available dynamic range, but yes, very compressed in the shadows. So adjust the dynamic range after applying this transfer function as aggressive halftone not affect the outcome. However, for disponéis still images taken with film, this method will allow you to take a much more linear aspect to the whole image, so I think that can be approximated quite what you get in peak condition with a DSLR and a single dose of 10-15 minutes.
Well then, in this early stage of processing, I think a good time to make a noise reduction, for two reasons:
- Apply a transfer function as aggressive tones media has left the bell-shaped histogram saw. That is, the noise is now everywhere distributed throughout the tonal range of the image.
- The algorithm for noise reduction and edge protection ACDNR adapts much more easily when the sky background values \u200b\u200bfound in this case in a higher range.
Fortunately, the read noise generated by the film scanner is relatively easy to remove by ATWT, as this type of noise is usually distributed at high spatial frequencies. So I just eliminated the first layer of wavelets, choosing a scaling function and a Gaussian kernel 3x3.
If however the noise generated by the film grain and is 'another story'. Here I had to apply two noise reduction:
- The first reduction includes a standard deviation value around 1.8 in 2.5 in the luminance and chrominance, using symmetrical threshold values \u200b\u200bfor edge protection. It is obvious that in these cases, a luminance mask inverted wavelets and curves adjusted, it always helps to define a much more controlled areas of the image where you must act more or less noise reduction.
- The second reduction is only for luminance, using a higher filter value (StdDev = 6.5). Also chose symmetrical values \u200b\u200bfor the protection of borders, but with much lower thresholds. Finally readjusted again the dynamic range, because as you know, the filter function in ACDNR tends to slightly reduce the contrast, especially in the areas of signal to noise ratio poorest, where the mask allows the filter to a greater extent .
Here is a good idea to also reduce chroma noise with SCNR in the green channel. Nothing new here to tell, except that in this case it worked better the method Maximum .
Now is the time neutralizes the sky. In this picture there are few regions can be defined strictly as a background, so I had to rely on running DBE interpolation after putting some samples at the top of the image. As simpre that successfully neutralizes the sky background, the left side of the hood of the histogram is more straight and steep, so that makes us free to leave a small portion of dynamic range in the shadows. So with HistogramTransformation again give him a light push on image contrast.
now the image is very far from showing a color balance. However, it is not possible to have a linear image because of reciprocity failure suffered by the film, is relatively easy to calibrate the color modifying the midtones independently on each RGB channel.
Finally, the most fun was to maximize color saturation by MTF on channel C (CIE L * c * h *) from LRGBCombination before taking action with multiscale processing, separating small-scale structures and removing them from the larger scales, where structures appear better defined in the Milky Way. HDRWT applied on these structures using eight layers, one iteration. As you know, this action increases the overall contrast to the entire image within defined structures. Then reinsert the result together with small-scale image of the original. This operation sometimes causes a slight imbalance between small structures and the rest of the image. So I took the opportunity to add more weight on small structures, while iteratively applied morphological and wavelet filters to the stars.
It is now a good gigs with CloneStamp .
