Renal Biopsy Image Segmentation Based on 2-D Otsu Method with Histogram Analysis

Renal biopsy image segmentation is an important step in computer-aided diagnosis. Among all available segmentation techniques, automatic thresholding methods are widely employed because of their advantages of simple implementation and time-saving. The Otsu method is one of these thresholding methods and is frequently applied in various fields. The two-dimensional (2-D) Otsu method is superior to the one-dimensional (1-D) method in segmenting images with a low signalto-noise ratio. However, satisfactory results are obtained only when the numbers of pixels in each class are close to each other. Otherwise, incorrect results are obtained. In this study, 2-D histogram projection analysis was used to correct the Otsu threshold. The 1-D histograms were acquired by 2D histogram projection in the x and y axes, and a fast algorithm for searching the extrema of the projected histogram is proposed based on the wavelet transform. The experimental results showed that the proposed method performed better than the traditional Otsu method for our renal biopsy specimens.