Digital Signal Processing Reference
In-Depth Information
2.6
Median Filtering
The median filtering is a nonlinear signal processing, the corresponding median filter
is a nonlinear filter. Linear filtering smoothly reduces the noise, but it blurs the image
edges and details. Nonlinear smoothing is to improve, that is, not all pixels with the
average of the field instead of a linear smoothing, but to take a threshold, when the
difference between the pixel gray value and the field average instead of at the thre-
shold when the mean; whichever is the gray value when the difference between the
pixel gray value and its field average is not greater than the threshold.
The median filter used in this paper is medfilter2 function. The value of each pixel
from the input image pixel values to determine the corresponding pixel in the field,
whose value is determined based on the median. The median sensitivity to outliers,
the median filter can remove these outliers and does not decrease the image contrast.
Usually median filter can filter out smaller hair noise, has a clearer distinction be-
tween the part of the hair bulb and none.
Fig. 7.
The image preprocess after median filter
3
Area Calculate of the Hairiness and Pilling
In order to better identify the fabric pilling, the degree of fabric wearing must be
ceded. The basis of classification is varied. In this paper, the ratio of the area of the
hair bulb and o the total area of pilling is used to decide the levels. In other words, the
proportion of the hair bulb area in the pilling fabric will determine the fabric hairiness
and pilling level.
In this paper the known fabric sample standard photo will first be used on in the
program, and each level of sample fabric pilling percentage values has been gotten,
This would be used to decide the test sample's level. Among them, the best pilling is
five, then four, three, two, the worst is one level. Then the ratio in the mathematical
way is used to calculate the value of the boundaries between the various levels to
determine the value of each level range.
3.1
Calculate the Hairiness and Pilling Part of the Area / Pixel
For the calculation of the hair bulb area, this paper is on the basis of the binarization.
After binarization, the image pixel values only 0 and 1. The function 'bwarea' is dedi-
cated to the calculation of the values of the two values of image pixels. Binarized
