Braille Cell Segmentation and Removal of Unwanted Dots Using Canny Edge Detector

Abstract

Braille cell segmentation is needed to detect, recognize and convert Braille cells of Braille document to natural language characters or word. The results can be used in the applications like converting Braille to text, embossing or printing Braille documents, distributing Braille documents over network or reproducing Braille documents on demand. Performance and quality of these applications get degraded due to noise introduced during scanning the Braille document. Noise may introduce due to scan resolution, poor lighting, introduction of skew or appearance of unwanted dots during scan. In line with the current work, skew correction and segmentation methods have been presented and published. Detecting the appearance of unwanted dots is a challenging task. The noise introduced during scan can sometimes be emitted as a dot in the scanned image. Also, deterioration of Braille plate can introduce some unwanted dots in the image while scanning. Poor lighting during scan also can be a cause of the unwanted dots. Identifying these unwanted dots is a major challenge. The relative position of the dot in Braille cell and their relativity with other dots helps to recognize Braille characters. Any unwanted dot makes it difficult in recognizing the character cell.