Skip to content
janrathert edited this page Nov 16, 2022 · 7 revisions

Welcome to the sr_stacker wiki!

1. converting the raspi raw images to pgm

read_raw *.raw it will output out_%06_r1.pgm , out_%06d_g1.pgm , out_%06_g2.pgm , out_%06_b1.pgm .

this is a single image of saturn . I took 250 images with 200ms shutter time . I used modified binoculars with one of the eye-pieces replaced by the raspi pi hq camera . saturn_single_picture

2. registering with select_star .

select_star -ref out_000000_g1.pgm select a point with the left mouse button . now there is a rectangle . you cam change the size with plus or minus on the numpad . press r to register the images . it will output out_000000_register_g1.txt this contains the positions and the difference of the sum images in frequency domain ( absolute values ) you can also display the image with a different blacklevel . select some region and press b to make the image brighter or darker press plus or minus ( not numpad )

3. drizzle

compute a larger image from many low resolution images

I used "drizzle -zoom 4 -square 2 register_g1.txt" to create this image

saturn_drizzled_with_zoom4

then I created a point spread function from a moon of saturn using select_star . the next image was created with cg_deconvol -lambda 0.5 -loops 50 -psf_pgm -o saturn_g_zoom4.pgm saturn_drizzled_g_zoom4.pgm . it was computed with the conjugate gradient algorithm .

saturn_with_deconvolution

Here I multiplied the matrix with the transposed matrix . and the equation system is A^TAx = A^T*b . it is better when the convolution kernel is not symmetric but has slower convergence .

saturn3 ed65b.png)

saturn2

4. bicgstab_deconvol

this algorithm works also if the matrix is not symmetric . the convergence is faster

moon_bicgstab