IDENTIFIKASI JENIS PENYAKIT KULIT BERDASARKAN ANALISIS WARNA DAN TEKSTUR PADA CITRA KULIT MENGGUNAKAN KLASIFIKASI K-NEAREST NEIGHBOR

Fitrianto, Faris and Isnanto, R.Rizal and Ajulian, Ajub (2011) IDENTIFIKASI JENIS PENYAKIT KULIT BERDASARKAN ANALISIS WARNA DAN TEKSTUR PADA CITRA KULIT MENGGUNAKAN KLASIFIKASI K-NEAREST NEIGHBOR. Undergraduate thesis, Diponegoro University.

[img]
Preview
PDF - Published Version
287Kb

Abstract

Humans have the ability to classify the image into the appropriate classes. These human capabilities when applied to software system, will be very useful. Therefore, research on human capabilities is required to be done which can be applied in the system. The purpose of the research is to develop a software that is able to classify the image into the appropriate classes using color and texture analysis, in this case, the image used of the diseased skin. In this research, several methods are used: using the color histogram, Edge Histogram Descriptor (EHD) or a combination of both of them. Level color histogram used 8, 64, or 256 optional colors. While EHD used 2x2, 3x3, or 4x4 optional region. The analysis begins with the color quantization process to obtain a minimum grayscale value for which its histogram is obtained. While texture analysis begins with the division of image into several regions and then five edge detection operators are applied: vertical, horizontal, 45 degrees, 135 degrees, and isotropic which the values will be determined as a block edge when exceeding predetermined threshold. These steps are applied to the training image which is then stored as database. The second step is to calculate the Euclidean distance between the values of color histogram and EHD on the image will be classified into an training image in database. The third step using KNearest Neighbor method is to determine image into one class of disease type. The highest level of recognition is achieved in combination method of color and texture features extraction, that in the 64 level of color histogram and 2x2 level of EHD which has an average accuracy of 68.57%. While the lowest level of recognition is achieved in 4x4 level of EHD method, that is 52.89%. This is caused by using combination of these methods, the extracted features has closer Euclidean distance to a particular type of disease rather than only uses one method. Keywords: color histogram, EHD, Euclidean distance, the skin disease

Item Type:Thesis (Undergraduate)
Subjects:T Technology > TK Electrical engineering. Electronics Nuclear engineering
Divisions:Faculty of Engineering > Department of Electrical Engineering
Faculty of Engineering > Department of Electrical Engineering
ID Code:32071
Deposited By:INVALID USER
Deposited On:20 Dec 2011 14:34
Last Modified:20 Dec 2011 14:34

Repository Staff Only: item control page