Video Scene Change Detection Using Mean Absolute Frame Differencing

Video Cameras are a common mode of monitoring behavior, activities and changing information in environments [1]. This change in information can be due to various reasons. These changes can be due to regular activities in the environment, like people moving around and utilizing space in the environments. Such activities do not require attention. However some of these changes could be of concern and need human attention. For example, a camera view intentionally blocked by a perpetrator could be a security concern, or a change in the orientation of the camera could be of concern.

This changing information due to any of the reasons is usually quantified by changing the intensity of the pixels in the image. Visually monitoring video cameras for these changes in tedious and intensive. Automatically detecting these changes could be of significant interest. The problems lies in identifying or classifying which of these changes in pixel intensities would corresponding to a reason that requires attention and which corresponds to those that do not.

Mean Absolute Frame Difference (MAFD):
A simple solution to this would be to track the change in the average pixel value of the image. If the change in the MAFD exceeds a certain threshold, a change in scene is detected.[2]

\[MAFD_n = \frac{1}{MN} \sum_{i=0}^{M-1} \sum_{j=0}^{N-1} |f_n(i,j) - f_{n-1}(i,j)|\]

\[M - Rows,
N - Columns,
f_n(i,j) - intensity-of-pixel (i,j) in-the-nth-frame\]

MAFD corresponds to the first-order derivative of $f_n$ and it measures the degree of dissimilarity at every frame transition.

MAFD can handle sudden changes in the scene but fails to detect gradual changes in scene. This is demonstrated in the following video.

The code for the above example can be found here.
https://github.com/pmantini/MAFD.git



Reference:
[1] Wikipedia
[2] Xiaoquan Yi and Nam Ling, "Fast pixel-based video scene change detection," Circuits and Systems, 2005. ISCAS 2005. IEEE International Symposium on, 2005, pp. 3443-3446 Vol. 4.