Adaptive Multiscale Block Compressed Sensing Algorithm Based on Total Variation

I191 multiscale block compressed sensing, the blocks within the same level are at the uniform subrate. However, considering different blocks have different importance for image reconstruction, we propose a total-variation-based adaptive sampling method for multiscale block compressed sensing. The algorithm obtains the total variation of the estimated image blocks via low frequency coefficients, then use the total variation to achieve adaptive allocation of subrate of the blocks within the same level, while it inherits the allocation of subrate of different levels from multiscale block compressed sensing. The experimental results show that, at the same subrate, the proposed algorithm outperforms some other previous work in terms of both the image reconstruction quality and the visual effect.