Mode determination in noisy bimodal images by histogram comparison

This letter introduces the "mode determination" problem relative to a bimodal image corrupted by noise, i.e., the problem of estimating the mean intensity level of each of the two homogoneous regions constituting a bimodal image. The mode determination problem can be seen as a generalization of the threshold determination problem. This generalization is introduced to cope with the low performances of thresholding techniques when dealing with highly noisy images. A mode determination method is proposed, based on the comparison of the histogram of different (averaging) convolutions of the given image. Experimental results are presented showing the performance of the proposed method.