Color RGB and Structure GLCM Method to Feature Extraction System in Endoscope Image for The Diagnosis Support of Otitis Media Disease

Abstract

This paper proposes an efficient technique for automatic detection of the tympanic membrane / eardrum in an endoscope image. All this time, the examination of the eardrum is done manually by a doctor to possibility for human errors. Then we need a system to help doctors diagnose the eardrum. The eardrum detection process involves four main steps. First, Preprocessing uses cropping and contrast enhancement to enhance lighting in the image, segmentation uses the grab cut model method to remove all non-eardrum pixels from the image, feature extraction uses RGB-color and GLCM to determine the value of color and texture features in the image and classification for determine the state of the eardrum. To handle ear detection of various ear shapes and sizes (triangular, round, oval and rectangular) and their size automatically adjusts the actual condition of the eardrum, without reducing the value of the image. System accuracy of 92.75 % classification accuracy by comparing the results of the system with the doctor's diagnosis, the RGB-color feature has the greatest effect compared to the texture feature. The proposed technique was tested on the eardrum image database of RSUD Dr. Soetomo consisting of 275 images of various kinds of eardrum conditions.