Script Identification of Multilingual Document Images Based on Block Finite Ridgelet Transform and Discrete Curvelet Transform
Zheng-Jian Wu, Reyihanguli Hasimu, Hornisa Mamat, Alimjan Aysa, Kurban Ubul · 2020
In recent years, many script recognition methods have emerged since they were studied as a front-end technique of OCR. These methods generally have a pleasing effect on a particular script, but they are not suitable for all languages. In this paper, we utilize the block finite ridgelet transform(BFRT) and discrete curvelet transform(DCT) and propose a fusion method in series for a total of 10,000 document images of 10 scripts including English, Chinese, Uyghur, Tibetan, Arabic, Turkish, Mongolian, Russian, Kazakhstan, Kyrgyzstan. The experimental results show that average accuracy is 99.35% in the classifier of linear discriminant analysis. Comparative experiments showed that the recognition rates of single BFRT and DCT were 89.03% and 86.3%, respectively. It demonstrates the effectiveness of the proposed method than the sole method. The validity of this method is proved by comparing it with some existing methods.