Collecting  Photons

Workflow of the processing of this M13 image. This processing is aimed at attaining an optimal sharpness of stars while stretching the differences in colour index between different star populations distributed on a wide dynamical range of magnitudes. This kind of processing should be quite useful for star clusters. All the displayed images are on the same scale (100%). Any comment is very welcome.

1) Preprocessing and separation of color and luminance channels Preprocessing is performed as usual (under Iris in my case) with offset, dark, flat and hot pixel treatment. The stack of 6x5' images (C8, f 6.3, Canon 300D, iso 400) is added and white balance is performed on a G2 star. The original image shows that the cluster is well resolved all the way to the centre with no star reaching saturation. In broad terms it appears that there are two classes of stars as distinguished by their colour index: white-blue stars (main sequence F spectra) and yellow-orange stars (red giant). Two copies of this stack will be processed separately to provide the luminance and chroma channels of the final image.

2) Luminance channels: stretching Luminance channel. The stretch is obtained with digital development (500, 4, sigma 1). The differences in colours are beginning to be washed out.

3) Luminance channels: deconvolution The sharpness of stars is improved by 10 iterations through a Richardson-Lucy deconvolution algorithm (LR 10). Noise is adaptively filtered (MMSE 160 10). The colour balance and saturation are severely degraded by this processing, but sharpness and contrast are clearly improved.

4) Chroma channel The colour information is maximized by operating an hyperbolic sine stretch (asinh 0.01 10 in Iris) of the original stack (step 1). This stretching maximizes the chromatic differences between the objects: specifically, it provides an optimal representation of the colour index of stars. Furthermore, it causes an adaptive gamma reduction, allowing a better display of the wide dynamic range of star brightness present in the cluster.

5) Export to Photoshop The two images (3 and 4) are saved in tif format and loaded in Photoshop CS. The chroma image is converted in Lab mode. In this modality the image is broken in three channels: L, a, b. The last two channels codify the colour info while the L channel hold only the luminance info. The Luminance image (step 3) is converted in a grey scale by averaging the green and blue channels of the RGB representation, the red channel is discarded since, in this example, it was noisier. On the left the image is split horizontally showing separately the original L channel of the colour image (above half) and the luminance images (below). The lower half is characterized by smaller star images and by a better definition.

6) Final composition. The L channel of the cLab image is replaced by the B/W luminance image. Sharpness is much improved maintaining colour saturation. Final adjustments of levels and small cosmetic touches and the image is ready. The final jpeg showed on page top is smaller than the original size, and received a small amount of unsharp masking (2 px, 65%).

Copyright © 2004 by Gimmi Ratto. (November 30, 2004)