In this paper, we present a novel adaptive method for ultrasound (US) image enhancement. It is based on a new measure of perceptual saliency and the view-dependent feature on US images. By computing this measure on an US image, speckle noise is reduced and perceptual salient boundaries of organs are enhanced. Because of the curvature gradient based saliency measure, this method can enhance more types of salient structures than the well-known saliency network method. Meanwhile, the proposed method does not depend on the closure measure. This makes it more appropriate to enhance real images than other existing methods. Moreover, the local analysis of speckle patterns leads a good performance in speckle reduction for US images. Experimental results show the proposed enhancement approach can provide a good assistant for US image segmentation and image-guided diagnosis.