The Application Of Denoising Algorithm Based On Document Image

The technology of the character recognition is an important mean that make the print words to a text that computer can recognize and could be edit by us. The traditinal methods to collect the document image is by the scanistor, but, the method which by the digital cameras to take photoes is gradually replacing the bulky scanning device. The new method is convenient, but also bring some new problem in the application, for example, the quality of the image may be affected by many external factors such as the uneven illumination and so on, these factors may have a significant impact on the character recognition rate. To deal with this problem,we use the image preprocessing before the character recognition. The method of image preprocessing firstly is to binarize, after the binarization,it may have residual noise, so the denosing is the necessary step, the common denoising algorithm contain median filtering, mean filter, mathematical morphology and so on, but these traditinal denoising algorithms can deal with any images, it has no targeted and the effectiveness and efficiency are not as welll as to deal with the document images.To deal with the noise in the document image that caused by uneven illumination in the process of image acquisition by camera, the architectural features and noise features of document images differ from the general images are studied, a fast denosing method through projection,block and morphology together based on the document image is proposed. Firstly,it is needed to binary the input document image.Secondly, after the binaryzation,it requires to project and get rid of the noise between the rows and columns based on the analysis of the projection. Lastly, according to the numbers of noise between lines and columns, a threshold is used to decide which block requires denoising,then the morphology is used to deal with the noise in the words. Experiments showed that the fast denoising method have improved the result and efficiency compare to the traditional denoising methods about the document image in uneven illumination.