Medical imaging is a useful technique for disease diagnosis and it has many applications in the medical field. There are several techniques used for medical imaging. Among them compression sensing (CS) technique has been widely accepted because of the low sample requirement and accurate recovery of image. In this paper, a novel adaptive matching pursuit for compressive sensing of blind sparsity biological signal polluted by noise is proposed. First, the traditional quadratic loss function is replaced with the more robust Huber loss function for the purpose of combating the influence of noise. Then, sparsity adaptive matching pursuit is introduced to make optimal estimation of the original biological signal and further reduce the influence of noise, thereby achieving accurate reconstruction of biological signal with blind sparsity. Simulation results indicated that the proposed algorithm greatly improves the anti-noise performance, especially in resisting large noise uncertainty compared with existing greedy algorithms.