Xiaohua Zhang, Jiawei Chen, Hongyun Meng, Xiaolin Tian
Optical Engineering, Vol. 51, Issue 12, 127001, (December 2012) https://doi.org/10.1117/1.OE.51.12.127001
TOPICS: Structural sensing, Image compression, Reconstruction algorithms, Compressed sensing, Principal component analysis, Image quality, Optical engineering, Autoregressive models, Matrices, Image restoration
A remarkable feature of compressive sensing is that the sensing is completely nonadaptive, but that does not mean that no effort whatsoever should be made to understand the signal acquired. With consideration of the differences and the similarities of the images, the proposed sensing matrix consists of two subsensing matrices: a fixed presensing matrix and a self-adaptive sensing matrix. The identical presensing vectors aim to estimate the compressibility of the image blocks and ensure the minimal measurements for reconstruction. The number of adaptive sensing vectors is flexible and depends on the compressibility estimated. More sensing vectors are assigned to incompressible image blocks than to compressible blocks. For a certain image block, the corresponding adaptive sensing matrix is constructed as a structured sensing matrix by combining the deterministic sensing matrix and the random sensing matrix. The former assigns the sensing vectors to the location where the information most likely lies and acquires the common structure contained in the image blocks, and the latter senses the difference and the individual structure the images possessed as an essential and unique characteristic. The proposed self-adaptive structured block sensing frame provides superior performance compared with many state-of-the-art compressive sensing reconstruction algorithms.