Documnent Image Binaration Based On Fuzzy Calculation

Binarization of documnent image plays an important role in preprocessing documnent image, the result of which will affect subsequent processing and identification. With the widespread of digital cameras and video cameras, acquisization of text graphics via digital apparatus in natural conditions has become an irresistable trend. Consequently, the technique of non-uniform and blurry documnent image processing has become very significant.This paper mainly discusses binarization of documnent image shot through digital cameras. Meanwhile, it further investigates the improved algorithms of the quality of binarized documnent image (noise removal, skew corrction).This paper summarizes some methods for documnent image binarization from numerous foreign and domestic scholars:global threshold binarization, local threshole binarization and the combination of the two. Besides, it puts forward two methods for documnent image binarization based on local threshold and fuzzy computation, and another based on global threshold and fuzzy computation. It also sums up their approaches for noise removal:spatial processing, frequency processing and mathematical morphology, and realizes noise removal on the basis of window scale and four-domain noise removal. In addition, it studys some methods for documnent image skew correction of domestic and internationed reseach, for example, Hough, Fourier transformation etc, and realize a skew correction based on text line tracking according to the characteristics of the document image strokes.This paper has collected a great amount of experimental data, proves the effectiveness of proposing and implementing new approaches, and improves the identification rate of characters during later periods. At the end of this paper, it makes a conclusion and points out the problems to be solved and the future development of document image binarization.