<p id="9z1vv"></p>

          <ruby id="9z1vv"></ruby>
            <ruby id="9z1vv"></ruby>

              <p id="9z1vv"></p>
              <ruby id="9z1vv"></ruby>
              <ruby id="9z1vv"></ruby>

                  ¡¡

                  Research

                  ¡¡

                                  Home        Research        Publication        People        Resources        News & Events        Forum
                  ¡¡
                      Current Topics
                  • Nonuniform filter banks

                    Multirate filter banks are used extensively in a number of signal processing applications such as subband coders for speech signals, frequency domain speech scramblers, and image coding. The theory of perfect reconstruction (PR) and near perfect reconstruction (NPR) filter banks with uniform frequency partition has been well established. However, in some applications, uniform filter banks are unable to provide appropriate time-frequency decomposition. For example, in the design of the filter banks that approximate the time-frequency resolution of human ears, nonuniform spacing that  matches the critical bands is preferred. This nonlinear arrangement is much useful when the signal energy exhibits bandwidth dependent distribution among frequency bands.

                    The main focus of our group is to analyze and develop efficient structures and effective design procedures for general nonuniform filter banks (NUFBs). In particular, we pay attention to those with linear phase (LP) property which is crucial in some applications such as image coding. In the recent work, Dr. Xie proposed an indirect method to design LP NUFBs with rational decimators based on the recombination structure. Then, a direct design approach of LP NPR NUFBs with integer sampling factors was proposed, where the transition bands of adjacent two analysis/ synthesis filters are forced to follow a cosine roll-off function to eliminate the amplitude and aliasing distortions. Later, the idea was extended to the rational case. Now we are woking on the design of M-channel irrational filter banks and another new method to obtain NUFBs with LP property.

                  • 2-D directional filter banks

                    Many multidimensional applications such as image enhancement, denoising, edge detection are germane to the directional information, so how to process this information is crucial. The directional filter bank has been proven to be effective in this field due to the attractive property that they can decompose an image into directional subbands£­each corresponding to a unique angular orientation, and reconstruct that image perfectly by using a dual synthesis filter bank.

                  • wavelet-like filter banks

                    The theory of the wavelet-like filter bank is just exploring. It may result in some compromise of the requirements of wavelets but at the same may present some useful properties in practical signal processing.

                      Projects
                  • The construction of M-band directional filter banks and its application to texture extraction of images

                    Supported by National Science Fund of China , Jan. 2007 ~ Dec. 2009.

                    Principal investigator

                  • The Theory and design of multi-channel directional filter banks and their applications

                    Supported by National Science Fund of Shaanxi, China, Jan. 2006 ~ Dec. 2007.

                    Principal investigator

                  • Design of time varying matched wavelet and its applications

                    Supported by National Science Fund of China, Jan. 2004 ~ Dec. 2005.

                    Participator


                  Page last modified: 2006-11-16

                    <p id="9z1vv"></p>

                          <ruby id="9z1vv"></ruby>
                            <ruby id="9z1vv"></ruby>

                              <p id="9z1vv"></p>
                              <ruby id="9z1vv"></ruby>
                              <ruby id="9z1vv"></ruby>

                                  ÒÁÈ˾þÃ×ÛºÏ