There are often many scars and hollows in ancient and famous trees. As a convenient and effective non-destructive testing tool, ground-penetrating (GPR) has a technical advantage in detecting abnormality in trees. But the tree radar images always inherit some extent of noise in them. Thus, denoising is very important to extract useful information from a tree radar image. Shearlet is a directional multi-scale framework, which has been shown effective to identify sparse anisotropic edges even in the presence of a large quantity of noise. This article presents an efficient denoising method based on shearlet applied on the tree radar images. Experimental results on forward modeling and standing trees radar data substantiate that the proposed method has the best denoising performance, especially in preserving the edge information as compared with the other methods which are based on wavelet, curvelet and contourlet.