Image compression using compressed sensing based on empirical mode decomposition

Based on bi-dimensional empirical mode decomposition (BEMD) and compressed sensing (CS) technology, an algorithm to improve image compression ratio is proposed. First, an image is decomposed using BEMD. As the amount of information at different layers is different and different layers have different sparse extents, therefore, according to the characteristics of different layers of BEMD, a method that using maximum feature points to construct surface mean is proposed. Multi-rate sampling is adopted to choose characteristic points, and the quality of the reconstructed image is also ensured. The first layer of the intrinsic mode functions (IMF) (IMF1) which contains more image details is retained, while other layers are compressed to enhance image compression ratio by using CS. For the reconstruction, IMFcould be recovered by using the orthogonal matching pursuit (OMP) algorithm. Then the image is reconstructed by interpolation. Simulation results demonstrates that, compared with the original BEMD compressing algorithm, the proposed algorithm requires less characteristic points to reconstruct image and improves the compression ratio from 3.7:1 to 4.7:1, at the same time, the reconstructed image's quality has no significant reduction.