Unsupervised Image Segmentation Using Textural Features
Önder Demir and Buket Doğan
Computer Engineering Department, Technology Faculty, Marmara University, Istanbul, Turkey
Abstract—In this study, a method was developed for the segmentation of the texture patterns in images containing multi-texture pattern. To distinguish these textures from one another, the image was divided into sections, and the textural features of each section were extracted using Gabor filters and gray level cooccurrence matrices. These features were analyzed using AGNES clustering algorithm and each section was assigned to a cluster. Segmentation was performed by marking as many different textural sections as the number of clusters formed on the actual image. The performance of the developed method was tested using the images obtained from Describable Textures Dataset database. The method made 37 correct segmentations for 40 images. In images with two different texture patterns, there were no incorrect segmentations. In 3 of the 32 images with three or four different texture patterns, incorrect segmentation was performed. Due to the sectioning of the image, borders between the texture patterns could not be precisely distinguished.
Index Terms—textural analysis, image segmentation, gray level coocurence matrix, Gabor filters, unsupervised clustering algorithms, image processing
Cite: Önder Demir and Buket Doğan, "Unsupervised Image Segmentation Using Textural Features," International Journal of Signal Processing Systems, Vol. 5, No. 3, pp. 112-115, September 2017. doi: 10.18178/ijsps.5.3.112-115
Cite: Önder Demir and Buket Doğan, "Unsupervised Image Segmentation Using Textural Features," International Journal of Signal Processing Systems, Vol. 5, No. 3, pp. 112-115, September 2017. doi: 10.18178/ijsps.5.3.112-115
Previous paper:Fusion of Dissimilar Features from Thermal Imaging for Improving Drunk Person Identification
Next paper:Last page
Next paper:Last page