Tuesday, November 4, 2014

First Mosaic

I wanted to capture the California Nebula which is huge! So, I tried for the first time a mosaic. SGPro's framing wizard makes capturing super easy.

You just draw a rectangle over the whole area that you want to capture and SGPro fills it in with overlapping images. The most important part is to have the rotation of the camera exact - otherwise the images will end up crooked. Creating a sequence of this mosaic creates two targets - one for each subframe.

I captured the images over several nights. Unfortunately, the vertical stripes were still in some of the images - I would try to get rid of them in post processing.

For processing, I first calibrated, stacked ... all images the same way in CCDStack.

For further processing, I found this tutorial how to do this with Pixinsight. I started to follow this.

First: cropping - nothing special here.

Resulting Ha images:

Next, we have to merge the images. First, using staralignment (Register/Union Mosaic). Strangely enough, Pixinsight could only find 30-40 stars in each image - which wasn't enough to overlap them. I had to play with various parameters in the "Star Detection" tab. Decreasing the "Log(sensitivity)" parameter detected more stars (40-50) but still not enough to match the two images. Increasing "Peak response" to 1 (maximum) lead to more the 100 stars being detected - which gave almost 20 stars in the intersection. Enough to overlap the two images:

The little black bars on the outside are because of cropping the individual images (and because the rotation wasn't 100% set correctly). But if we look closely at the intersection we can see the seam from the two images - this will be tough to remove. That's why Pixinsight has another process to smoothly stitch mosaics: GradientMergeMosaic. It's a little tricky to apply (see tutorial) as it requires a few temporary files to be created. But after doing all that, I could finally apply it and got this:

There is clearly something wrong at the seams. After closely inspecting and trying out a few things, I realized that this is a result of the different background levels of the two images. So, we need to remove the background gradients of the individual images first (DBE). Doing that* and then using GradientMergeMosaic results in this:

Even under close inspection, there are no seams left: 

From here on it's normal processing. Including removing gradients (once more) in the combined images. The reason is that the two images still had slightly different background levels and combining them resulted in a (smooth!!) gradient.

*Also, by removing the background gradients first, I could use the default star detection parameters in StarAlignment!