Intravascular ultrasound (IVUS) has been applied into the diagnosis of coronary heart disease and the relevant intervention therapy more and more widely. However, the frequency of ultrasound must be increased in order to improve the resolution of images, which makes the blood speckle noise increase significantly,decreases the contrast of lumen and arterial wall structure, increases the difficulty to discriminate arterial wall from structures, and makes it inconvenient to diagnose and treat illness. In this paper a wavelet shrinkage method based on local threshold is introduced to reduce blood speckle noise. Results show that the method can remove the speckle noise, keep the edges of image, increase the contrast, and thus is helpful to the identification of the arterial wall and the tissues around.