A new symmetric extension of the signal at the boundaries for multiwavelet image compression

The paper proposes a new symmetric extension of the signal at boundaries for multiwavelet image compression. In practice, all image signals are of finite length, so they should be extended at boundaries before being processed with wavelets or multiwavelets. The most common way is periodization of the data, but it causes discontinuities at the boundaries that are not what we want. Symmetric extension of data can resolve this problem. The classical symmetric extension for GHM (Geronimo-Hardin-Massopust) multifilters (multiwavelet filter banks) requests that one input signal is even-length and the other is odd-length. Also, the quantity of the high frequency output data are more than the quantity of the low frequency ones, which is disadvantageous for data compression. We address a new algorithm to resolve these problems. The result of the experiments indicates that the energy distribution properties with this new algorithm are 6.3% more than those with classical symmetric extension and 11.8% more than those with periodization extension.