Adaptive binarization with non-uniform image illumination

Abstract

The article considers the image binarization process for its further recognition. During binarization of the image with non-uniform illumination some noise appears which makes it impossible to select features in the image. Considering the QR-code image as an example, Otsu method, Niblack method and Bernsen method were analyzed. Some errors in binarization of the image with non-uniform illumination were detected in the process of application of these methods. Therefore, the necessity of finding a better method arose. The presented method of adaptive binarization is a simple extension of Bernsen method, the main idea of which is the comparison of each pixel with arithmetical mean of pixels in its environment and simultaneous application of integral presentation of the image.