Digital Signal Processing Reference
In-Depth Information
Fig. 7.
A milk bottles image and its automatic segmentation of scale area: (a) Projection image;
(b) Smooth the results; (c) Scale segmentation; (d) Non-scale area
Hx
()
Hx
'( )
Hx
''( )
Fig. 8.
The gray level histogram of the sub-region
Fig. 9.
Smoothing the histogram, the
first derivative and second derivative
by Shen algorithm
In Fig.8, the rightmost peak corresponds to the region of the grayscale higher milk,
while the left peak corresponding to the sidewall part. Milk area is split by finding the
valley value between the right side of the peak and the left peak point as an adaptive
segmentation threshold. In fact, the valley point position is seen as a minimum point. But,
it is difficult using conventional mathematical methods to calculate the histogram of the
first derivative and second derivative. However, in the above, we can easily calculate the
derivative is 0 points and to determine the derivative of the positive and negative through
the introduction of Shen algorithm. Then adaptive threshold is found by a valley point.
Fig.9 shows the sub-image histogram H(x) by Shen filtering and its corresponding
first derivative and second-order derivative (H'(x), H”(x)).
In Fig.10(a), we can see that grayscale image is converted into binary image by
threshold segmentation technique. However, the binary image due to background
noise, and some small noise region, which affects the location of the accurate posi-
tioning surface. Therefore, these noise region needs to be cleared.
4.2
Region Marking Method for Noise Region
Fig.10(a) shows milk part is the largest in all white areas. Therefore, the largest area is
found by the regional marking method. First, each region is marked through regional
