Phase noise cancellation for digital holographic microscopy based on compressed sensing iterative adaptive sparse dictionary

Abstract

Digital holographic microscopy (DHM), like other digital imaging techniques, will introduce severe coherent noise, which is called random noise, due to the high coherence of the required light source and the measurement experimental environment. Random noise seriously affects the accuracy of DHM surface topography measurements. To overcome this problem, this paper proposes a DHM phase noise cancellation method based on compressed sensing (CS) iterative adaptive sparse dictionary. The holographic image is first acquired using the proposed recording method. Then a sparse dictionary more suitable for the sample is trained by dictionary learning to sparsely encode the holographic image. Finally, reconstruction is performed using the improved orthogonal matching pursuit (IOMP), which can obtain almost noise-free holographic images. The experimental results show that the reconstructed phase peak signal-to-noise ratio (PSNR) is 33.2871, which is improved by 65.01 %. The sample feature similarity is 0.9553, which is improved by 45.41 %. The proposed method can also be used with different learning modes and learning objects. On top of the original results, the reconstructed phase PSNR was able to continue to improve by 5.99 % and its reconstruction time was also accelerated by 29.57 %.