Background Correction using Average Filtering and Gradient Based Thresholding

Abstract

Segmentation process on the image with illumination and contrast variation problem is a very challenging task. This problem can reduce the effectiveness of segmentation result. Therefore, the implementation of the proposed method based on the background correction is able to improve the image quality and automatically increases the segmentation. The proposed method used in this study is based on mean filtering and Otsu thresholding techniques to enhance the non-uniform image for better segmentation. The proposed method used the mean value of the image to normalize the background image. Then, the resulting image from the previous step underwent the segmentation process using Gradient Based Adaptive thresholding. Finally, a comparison in term of misclassification error (ME) was calculated and compared with the six other methods. For the ‘rectangles’ image, our method with gradient achieved 0.050478 and it is better compared to the other six methods. However, the ME value of the ‘text’ image produced by our method is 0.058722, slightly higher than the Niblack’s method, Chen’s method and gradient based method. Therefore, it still acceptable in comparison to those methods by Yanowitz and Bruckstein’s (YB) method, Blayvas’s method, and Chan’s method. The proposed method is better method to enhance and improved the image quality. The main impact of this study is to eliminate the illumination and normalize the contrast variation. In conclusions, the implementation of the proposed method produces an effective and efficient results for background correction and increases the segmentation result.