When using ultrasonic time-of-flight diffraction (TOFD) to detect defects in titanium alloy welds, due to the coarse grains, anisotropy of titanium alloys and the complex composition of welds, TOFD images have large noise and small differences between defect region and backgrounds region. At the same time, the existing preprocessing methods cannot balance the contradiction between noise removal and defect feature retention, which seriously affects the accuracy and effectiveness of defect recognition. In order to overcome the above problems, an anisotropic diffusion model (P-M) based on region adaptive strategy is proposed to realize TOFD image enhancement. Firstly, by analyzing the change of information entropy in TOFD image, a segmentation method of defect region and background region based on information entropy is proposed. Secondly, an adaptive selection strategy of diffusion coefficient in P-M is established to achieve adaptive noise reduction and detail enhancement in different region of TOFD images. Finally, a case of TOFD image enhancement is given to illustrate and validate the proposed method. The results show that the proposed method can effectively remove the noise of TOFD image and preserve the defect details compared with the traditional filtering method.