Abstract: In reconstructing acoustic object and predicting sound scattering based on the Fourier diffraction theorem,projections in all directions are calculated by rotating the object while maintaining the incident wave unchanged.Object rotation is time consuming,and suffers from errors due to ruggedness of the edges and holes inside the object caused by dicretization.To improve computation efficiency and quality of the rotated image,the discretized binary object in the Cartesian system is converted into polar coordinates and presented on rectangular grids.In this way,rotation of the object is achieved by translation in the rectangular grid-represented polar system,and the result is converted back to the Cartesian system.The obtained image is then smoothed with morphological opening and closing operations.The proposed method can significantly reduce the computational burden,avoid holes and reduce edge ruggedness,thus improve accuracy of the computed projections.