As a part of my project I had sliced short B&W video in frames, then denoting each frame as a matrix I reshaped 3D matrix ([frame]x[vertical_pixels]x[horizontal_pixels]) into 2D matrix - [frame]x(vertical_pixels * horizontal_pixels). After that I applied singular value decomposition (SVD) and sorted $\Sigma$ main diagonal elements in descending order. Plotting all that singular values got me a following graph (Log scale): .
And now I'm confused by min/max values being detached from the main trend. Do you have any ideas why it is happening?
EXTRA: It is CCTV recording, the road is being recorder and several people are passing by.
