Image Analysis on stars/streaks in sky

I'm working on a program that will find out the orientation and location of streaks in the sky are based on images taken from telescopes in order to update current skycatalogs on where orbital debris and satellites are. Currently I have determined what pixels in the image are part of a streak (bright white lines), and I have put the horizontal and vertical indices of these pixels in a 2-d array. In (example image), the left side is the pre-processed image and the right side I have replaced all the pixels in the streak with the average noise value of the image. Exampleimage2 shows another example input with an obvious streak.

My question is, how can I determine the width of the streak if I already have the horizontal and vertical indices of all the pixels in the streak (note: the width would be the shorter dimension, probably only a few pixels wide).

Answer 1

[Don Zheng]

Try Principal Component Analysis and the width should be proportional to the standard deviation of the second principal component.

MATLAB provides this ' pca ' function to help you do that.