Orthogonal transform single-pixel imaging with fewer measurements

Orthogonal transform based single-pixel imaging methods always improve the quality of recovered images, but they should take two-fold measurements of the illumination pattern the under full sampling. To decrease the number of measurements for time efficient imaging, we present a method that only uses the positive patterns to acquire measurement values and reconstruct images, the number of measurements can be reduced by 1/2. The robustness of the method is guaranteed by eliminating the random noise with the characteristics of orthogonal transform series. Simulation and experimental results based on discrete cosine transform single-pixel imaging, Hadamard transform single-pixel imaging and discrete W transform single-pixel imaging validate the effectiveness of the proposed method.