An investigation into sub-band adaptive filtering of digital images

Digital images can become corrupted from various sources including sensor noise, communication channels and atmospheric interference. Signal processing techniques have been used in an attempt to improve the signal quality. Single-channel frequency domain filters (e.g. lowpass, highpass) cannot discriminate between signal and noise when their spectra overlap, thus signal distortion may be severe. Sub-band adaptive filtering (SBAF) has displayed advantages over full-band adaptive filtering (FBAF) when applied to speech processing and echo cancellation, and preliminary results by Smith and Campbell (see Proceedings of the fifth IEE International conference in Image Processing and its Applications, p.722-26, 1995) suggests benefits can be gained by applying SbAF to the two-channel image restoration problem