The description of directional information in image is always an important issue in image processing. There are many methods to extract the directional information of images. However, the underlying filter banks are often complicated in design and the frequency tiling suffers from some limitation in some extent. This paper proposes a new directional filter banks based on DFT modulated filter and applies the new directional filter banks in image denoising. The directional filter bank (DFB) based on 1-D DFT modulated filter is different from separable wavelet filter banks. It avoids the directional across in diagonal subbands and provides finer directional information than wavelet filter bank does. It is also different from the conventional DFBs and the new DFBs are more flexible in directional selectivity. Further, the proposed DFB is applied to image denoising and the experimental results show that the denoising algorithm using the new DFBs can significantly improve denoising performance.