Ore image thresholding using bi-neighbourhood Otsu´s approach

A new thresholding algorithm based on two neighbourhoods and Otsu´s method is proposed. For each pixel, its two neighbourhoods are chosen by image and object sizes. The thresholds of neighbourhoods are optimised by Otsu´s method, and then every pixel is binarised through the minimum of thresholds. Experimental results show that the proposed algorithm improves thresholding performance efficiently, especially for multi-object and uneven lighting images.