Mobile phone cameras and still cameras nowadays. There is a very sharp appreciation is down and you can easily carry it from the camera too low quality to be used with programs recognized characters (OCR) because the light in the shadow of the camera focus photography and other factors. From the observation of unique images from scanners, digital cameras, laksa, I have proposed a new Binarization method, which is used to split the image region as subsections to the complexity in each region are less pictures. Using the camps as Coeffcient Skewness After that, where there are no characters are excluded by using the method of Difference of Gaussian best remaining area will be by means of the Otsu Threshold was adjusted to be able to work with images that have dark letters on a light colored background as well. As well as images that contain a light on dark background. The proposed method was applied to compare with other methods such as Niblack Hui-Fuang how to's, and the document is saved with the condition by Sauvola digital camera at 72 dpi resolution, pixel size and short focal 1600x1200 120. From the experiment found that the error value from the average letter recognition of how I have this presentation to 14.59% which is higher than the Sauvola method is the best way to experiment with just 4.92%, but how can this be developed faster and simpler because it does not require the user to complete the process.
การแปล กรุณารอสักครู่..
