自適性閥值運算與應用條件性連通物件於影像文字辨識

Apply Adaptive Threshold Operation and Conditional Connected-Component to Image Text Recognition

在文字辨識的領域裡，清楚且正確的將文字從圖像中擷取出來辨識，是個非常關鍵的議題。經過影像前處理後，擷取出的字詞完整與否將影響著文字辨識的準確率。圖像中的文字是人們感興趣且具有意義的一部分，但一張圖像中往往存在許多干擾元素，如不同的光線強弱或複雜的背景，而這些元素常常會增加字元辨識的困難度。在本研究中，運用自適性閥值運算以及機率性連通物件來解決不同光線的強弱和複雜的背景所造成的影像問題。實驗結果顯示，經由本研究影像前處理方法，在字元辨識的結果中，正確性有著顯著提升，文字辨識率與識別率分別高達81.17%與91.30%。

How to effectively extract texts from an image is a critical issue in text recognition domain. After image preprocessing, the wholeness of the extracted text region will profoundly affect the accuracy of further OCR processing. A well-planned image preprocessing is believed to produce better OCR results. Due to the variety of background components, for example, different kind of colors, texture, or brightness in an image will deteriorate the problem of text recognition. In this research, we applied “conditional connected-component” and “adaptive threshold operation” to deal with complicated background and non-uniform lightness images. With this approach, we successfully identified and recognized texts from an image. The result shows that the rate of object identification and recognition achieves 81.17% and 91.30%, respectively.