Try processing the input image with Image Magick morphology methods. Detailed examples are here. The script below will size down and threshold the image to 2bpp bitmap, then it will try to close all the gaps, and round the shape. Afterwards a skeleton is created that holds general information about our shape. Finally using hit and miss kernel you search for specific line ends. In this case diagonal ones. The output is an image with set of dots. However you can use Image Magick's identify
to print pixel positions. I also assume that more or less you know where shoulders are and can discard points in other parts of the image. The C-point, identifying neck can be extrapolated from A and B.
convert -resize 25% input.jpg -negate -normalize -colorspace gray -threshold 20% -trim +dither v.png
convert v.png -morphology Close Disk x.png
convert x.png -morphology Thinning:-1 Skeleton:2 y.png
convert y.png -morphology HMT 'LineEnds:2;LineEnds:2>>' z.png

I used IM v.6.7.9 as I have some issues with v7.