2U-4
A Proposed Multi-Feature Extraction Method for Khmer OCR
○ワンナ クルイ,亀山 渉(早大)
OCR technology has already matured. However, there is no reliable OCR system for Khmer Language. This is largely due to the lack of Khmer OCR research efforts and the complex nature of Khmer characters. Most of Khmer OCR researches focus on single feature with limited accuracy. We propose a method based on multi-features using Scale Invariant Feature Transform, Fourier Transform Descriptors, Holes and their positions, and component’s location. In this paper, we present our results, considerations, and future work.