The update method is measuring the instability of the
pixels; the pixel may be the motion pixel or the noise if it
is unstable, on the contrary, the pixel will be updated if it
is stable. An important thing to note here is that the
measured instability of pixel is the mask area of motion
excluded, in other words, it is the pixel of non-motion
area. These pixels are likely to belong to the background
but it does not mean that these pixels of non-motion area
are equal to the background. We need to measure the
stability of these pixels in order to determine whether
these pixels belong to the background. The measurement
function is to measure the degree of difference between
the pixel values in consecutive frames in the same time
and same place. When the gap reaches to a definite level,
it represent the degree of difference is too high. Then we
have to re-measure the real peak-value and valley-value to
replace the older background value. The following
inference is made: ( ! represents the peak -value of
background, " represents the valley-value of background
and #
represents the stable pixels). ! and " of
the timeliness background is updated in (4):