K. Engan and K. Skretting (Norway)
overlapping frame, denoising, matching pursuit, dictionary
The use of frames and matching pursuits techniques for
signal representation are receiving increased attention
due to their potential in various signal processing appli
cations. Good design algorithms for block-oriented and
overlapping frames have been published [1, 2, 3]. In this
paper we demonstrate a novel technique for denoising
of images. The technique uses overlapping frames and
Fully Orthogonal Matching Pursuit (FOMP) as in ,
in conjunction with soft thresholding. Experiments
are done on image with additive Gaussian noise and
on JPEG compressed images. The experiments are
compared with wavelet denoising and median ﬁltering.
The results shows very good potential for denoising
using overlapping frames (dictionaries).