Iterative hard thresholding (IHT) algorithm is one of the representative compressive sensing (CS) reconstruction algorithms. For applying to images, however, it has a problem of lacking in addressing human visual system (HVS) characteristics - its hard thresholding process treats all of coefficients in transform domain equally. To overcome the problem, this paper addresses an adaptive hard thresholding method accounting for the HVS characteristics. For this purpose, a suitable threshold level is adaptively selected for each coefficient in transform domain by utilizing the standard weighting matrix table used in JPEG together with the threshold value which is estimated over the noisy version of image. Experimental results show that the performance of the block compressive sensing with smooth projected Landweber (BCS-SPL) with the proposed adaptive hard thresholding algorithm remarkably outperforms that of the conventional BCS-SPL algorithm. © 2013 Springer Science+Business Media Dordrecht(Outside the USA).
CITATION STYLE
Nguyen, V. A., & Jeon, B. (2013). Recovery algorithm for compressive image sensing with adaptive hard thresholding. In Lecture Notes in Electrical Engineering (Vol. 240 LNEE, pp. 949–955). https://doi.org/10.1007/978-94-007-6738-6_117
Mendeley helps you to discover research relevant for your work.