Thresholding is an important technique for image segmentation that extracts a target from its back-ground on the basis of the distribution of gray-levels. One of the popular automatic threshold selection meth-ods, the Otsu method, provides satisfactory results for thresholding images with obvious bimodal gray-level distribution. However, Otsu method fails if the foreground and background pixels have significantly different variances, or if the histogram is unimodal or close to unimodal. Valley-emphasis method partially resolves such problems by weighting the objective function of the Otsu method using the probability information at the valley-point in the histogram. In this study, a new weighting scheme is proposed for improving Otsu threshold selection which incorporates a measure of valley deepness with probability of occurrence at the threshold location to enhance the weighting effect. Experimental results indicate that the proposed method greatly improves the performance of the Otsu method and is highly competitive with other widely used thresholding methods.