| Summary: | 碩士 === 華梵大學 === 電子工程學系碩士班 === 94 === The shapes of images contained characters, numbers, fingerprints, or the license plates, etc. are all geometric. Based on the basis frequency image and the analytic properties of the two-dimensional frequency domain image for the geometric image, a new asymmetric filter is proposed in this paper. It can be used to fetch the messages of horizontal edge, vertical edge, bevel edge, and etc. in a geometric image. Therefore, it can be introduced to many applications, such as the classification of images between pictures and texts and the classification of text images between different languages. In this paper, we apply the asymmetric filters on the classification of images between pictures and texts and the classification of text images between Chinese and English languages. Then we get the nice successful percentage of these classifications. To review the traditional OCR (optical character recognition) system, since the characteristic values and the methods used in OCR are different for different languages, user must select what kind of the language used before OCR. The proposed method will be used to replace the artificial manually select-mode in OCR. That will lead us to more convenient using on the interfaces of human and devices.
|
|---|
