Depth-of-field reduction due to blurring in a relayed plenoptic camera and mitigation via deconvolution

Abstract

The projection of an externally-mounted microlens array (MLA) onto a sensor via relay lenses is oft adopted by the scientific community in lieu of traditional plenoptic cameras with embedded on-sensor MLAs. In most cases a relayed design better supports prototyping and temporary setups, while in other cases such as intensified plenoptic imaging the extended projection distance is necessary due to physical constraints. However, relaying the MLA inevitably blurs the raw image, often by 1.5 px or more. We observed an unusual behavior where fixed blurring at the raw image level produced blurs that are worsened by depth within refocused images – enough to reduce the plenoptic system’s depth-of-field by 60% in most cases. To address this problem, we hereby put forth a model to elucidate the mechanism of this depth-dependent blurring, and propose a Lucy–Richardson deconvolution procedure for mitigation. Discussions are undertaken in the contexts of regular and intensified plenoptic photography, as well as application in 3D particle image velocimetry for fluid dynamics. Results were validated on synthetic as well as experimental images.