COMPRESSED SENSING BY ITERATIVE THRESHOLDING OF GEOMETRIC WAVELETS: A COMPARING STUDY

摘要

Recently, a new compressed-sensing (CS) theory for simultaneous sampling and compressionnof signals has been applied for imaging and remote sensing. The CS makes itnpossible for us to take super-resolution photos only using one or a few pixels rathernthan millions of pixels by conventional digital cameras. However, the performances ofnCS are related to choices of a measurement matrix and a sparse transform. In thisnpaper, we present an experimentally comparing study for the use of different measurementnmatrices (e.g., random matrices, noiselet transform matrices, and scrambled blocknHadamard ensemble) in encoding step and different geometric wavelets (e.g., curveletsnand bandlets) in decoding step. Numerical experiments for single-pixel imaging andnFourier-domain multiple-pixel imaging indicate how to choose a suitable CS strategy tonreduce the number of measurements and decoding costs