Alignment and disruption robust binary mapper for optical Braille recognition

Abstract

Braille is the system invented for visually impaired peoples to capable them for written communication. The basic constructs of each natural language can be transited to corresponding Braille characters. Optical Braille recognition system is capable to read the Braille pattern images and convert it to corresponding language text. The images captures from portable camera devices and scanners are generally geometrically and visually disturbed images. In this paper, a disruption robust model is presented to improve the scope and accuracy of Braille character recognition system. The proposed framework improved the image visibility using SD adaptive filter and used the polar cosine transformation to resolve geometric misalignment. Later on grid mapper and binary encoder are applied on emboss enhanced character image for transforming it to binary coded array. Finally, the peered map on this binary character is implied to recognize the corresponding language text. The proposed framework is implemented on three different sample sets of Braille character and document images collected randomly from the web. The comparative evaluation against PCA (principal component analysis) and LDA (linear discriminant analysis) method shows that the proposed model significantly improved the accuracy of Braille character and document recognition.