One of the key issues in ultrasonic tomography is the calculation of sensitivity and image reconstruction based on the sensitivity. A method is developed to calculate the sensitivity of ultrasonic tomography in rectangular area in this paper. On this basis, numerical modelling calculation of different substances distribution in a rectangular area is made of by using LBP algorithm, the standard Tikhonov regularization method and the Landweber iteration method respectively. Reconstruction simulation experiments of changing the quantity of pixel is conducted. The results show that imaging results of Tikhonov regularization method and the Landweber iterative method are consistent with the original image for the overdetermined problem and superior to those reconstructed by LBP algorithm. The image quality of Tikhonov in the underdetermined problems is poor than that in overdetemined problem and still better than that in the Landweber and LBP algorithm. Compared by numerical simulation, Tikhonov regularization method in both overdetermined or underdetermined problems are better than Landweber iteration and LBP algorithm, and LBP algorithm is only suitable to qualitative analysis.