Enhancing digital holography through an AI-powered approach for speckle noise reduction and coherence length preservation

Digital holography (DH) has become a promising tool in various research fields for acquiring quantitative phase information (QPI). However, its reliance on high-coherence light sources such as lasers often leads to speckle noise, which degrades image quality. Although low-coherence sources like light-emitting diodes (LEDs) can mitigate this noise, they struggle to create complete interference patterns for specimens with optical path differences exceeding their coherence length. This trade-off between high coherence and low speckle noise presents a significant challenge in DH, particularly in applications requiring long coherence lengths for accurate QPI. Our research addresses this challenge with an AIpowered approach. By training an AI model with paired hologram data from lasers and LEDs operating at the same peak wavelength, we have developed a method to reduce speckle noise while preserving the coherence length. The newly proposed method has been verified on reflective specimens using a Michelson interferometer. The resulting holograms from this AI model exhibit clear interference patterns over depths that match the laser's coherence length, while simultaneously achieving significantly reduced speckle noise, akin to that observed in LED holography.