Segmentasi citra tanaman hias menggunakan metode Boykov and Kolmogorov max flow/min cut graph

Abstract

Segmentation is an important step to improve pattern recognition. But until recently, few studies have been done yet to segment the image with complex background. This research is trying to segment the image that has a complex background. This study used Boykov and Kolmogorov max flow/min cut graph for segmentation. This method uses all the pixels to form a directed graph with two terminals. Database of 300 house plant images belong to 30 different types of house plant in Indonesia. In testing, 240 images are used and then extracted using Rotation Invariant Uniform Patterns () and performed recognition using probabilistic neural network (PNN). Results showed that accuracy increased between 8.33% to 22.22% after the segmentation.