Adaptive acquisitions in biomedical optical imaging based on single pixel camera: Comparison with compressive sensing

Single pixel imaging opened the door to a cheaper camera architecture able to operate in a wide spectral range. Such an optical setup has been used with compressed sensing to reconstruct an image via ℓ1-minimization ruling out real time applications. In order to have a direct restoration of the image, we consider an adaptive approach for which we propose a new acquisition strategy. Our method progressively acquires an image in the wavelet domain by predicting the significant coefficients. For this, we base our technique on the non-linear approximation of the wavelet transform taking advantage of the transformation's sparsity. This new strategy is shown to offer high performance on simulated and real data that we compare to compressive sensing acquisitions. One possible application of the single pixel camera can be foreseen in fluorescence images of biological structures.