Open AccessThe outlier pixel and sub trees that form segments of image
Author(s) -
Efron Manik
Publication year - 2020
Publication title -
iop conference series. materials science and engineering
Language(s) - English
Resource type - Journals
eISSN - 1757-899X
pISSN - 1757-8981
DOI - 10.1088/1757-899x/725/1/012125
Subject(s) - outlier , segmentation , pixel , pattern recognition (psychology) , artificial intelligence , mathematics , image segmentation , image (mathematics) , standard deviation , sample (material) , threshold limit value , anomaly detection , computer science , statistics , physics , medicine , environmental health , thermodynamics
Segmentation is a very important step in computer vision. Many segmentation methods are found to get the best results. Every method found always has advantages and disadvantages. Minimum spanning tree is one method used to segment images. Some researchers use thresholds as linear combinations of averages and standard deviations for this method. This study aims to see the effect of outliers on the number of segments formed. This study uses a sample of a 5x5 image with an outlier. Suppose that the distribution of side weights (or the difference in intensity of adjacent pixels) is a normal distribution except for one outlier, and the threshold value is equal to the average. If the weight of the outlier data side is equal to 14 to 16, there are 6 segments formed, for weights equal to 17 to 37, there are 4 segments formed, and for weights equal to 38 or more, there are 2 segments formed. The threshold value (corresponding average) changes significantly by changing the value of one outlier data. This will cause changes in the number of segments in the image that are not good in the segmentation method. So the threshold value should not correspond to statistical values.