A denoising method for phase-retrieval-based imaging

Abstract

Denoising is significant in many fields, especially for computational imaging. Coherent diffraction imaging and speckle correlation imaging are regarded as the most promising computational imaging techniques. The above two imaging techniques can be classified as phase-retrieval-based imaging due to the phase-retrieval is a vital procedure for object reconstruction. However, the acquisition process would generate unavoidable noise and participate in the iteration process of phase-retrieval. Hence, it is necessary to denoising after obtained the original reconstruction image. Here, a denoising method that based on connected domain is proposed for phase-retrieval method. We experimentally demonstrate the denoising results and quantitatively analyze the effect. Comparison of the classical median filter, wiener filter and bilateral filter, our method shows a satisfactory denoising effect. Our results prove that connected domain denoising is useful and promising, which provides a new post-processing denoising method for phase-retrieval-based imaging.