Segmentasi Citra Tanda Tangan Menggunakan Fitur Titik SURF (Speeded Up Robust Features) dan Klasifikasi Jaringan Syaraf Tiruan

Signature image classification is an important field of image processing. One of the stages of signature classification is segmentation. The segmentation process aims to detect image pixels that are part of the signature and separate them from text or logo pixels in a document image. There is a signature segmentation technique using interest points extracted using the SURF (Speeded Up Robust Features) algorithm [1] In this technique, a connected component pixel will be considered part of the signature if it has more SURF points in common with the database connected component pixel signature. Compared to the similarity with the database connected component non-signature pixels. This method is able to provide good accuracy results for signature pixel segmentation. However, the recall value is relatively low, namely 56%. This is because many connected component logos are considered as connected component signatures. In this study, signature segmentation was carried out using SURF points by adding two things: 1) using internal connected component characteristics as additional classification atributs: extent, solidity, ratio, and circularity 2) using an Artificial Neural Network classification algorithm to assist the segmentation process. The test results show that the proposed method improves segmentation quality by an average of 20.7% for an increase in accuracy, an average of 22.4% for an increase in precision, and an average of 18.6% for an increase in recall. When compared with the results reported in (Ahmed et al., 2012), the recall has increased by 38.3% - 42.8%