Differential sparse single-pixel imaging with wavelet coefficients

Wavelet tree-based adaptive compressed ghost imaging has higher imaging speed and better imaging quality than compressed sensing methods. However, the number of samples can be reduced, since finer significant coefficients can be calculated with the existing measurements. This paper proposes a compressive adaptive ghost imaging with block-wise computational sampling in this paper. This method takes full use of the sparsity of the image in the wavelet domain. The block-wise computational sampling computes four LL band coefficients of high resolution via fewer than two measurements. The calculation strategy takes the useful information of the sampled low-resolution coefficients. This brings in fewer measurements. The computed LL band coefficients are also used to sample significant HH, HL, and LH coefficients of the same scale. The extended prediction method can accurately locate the positions of the important coefficients, by eliminating insignificant coefficients that belongs to significant parent coefficient. The simulation results show that the proposed compressive adaptive ghost imaging using block-wise computing sampling can achieve lower rate of sampling and advanced imaging quality.