Compressive sampling of natural images using bandelet basis

Bandelets transform exploits geometric information of image to determine bandelets coefficients. Bandelets basis has been used for compression of geometrical images and image denoising. In this paper, bandelets basis are used for compressive imaging of natural images using benchmark and efficient l1 minimization algorithms. Conventional compressive sampling matrices would not be suitable to extract geometric information required for construction bandelets basis. In this work, a so-called regular point sampling matrix is used to find out the geometric regularity of images. The experimental results on a set of natural images show that natural images can be reconstructed by using bandelet basis with higher PSNR and good visual quality. The bandelets based compressive sampling method outperformed conventional compressive imaging frameworks.