A data domain information measure for incoherent computational imaging

The developments in the area of computational imaging over the last three decades are aimed at increasing the information gathering capacity of imaging systems. This enhanced imaging performance is achieved by use of unconventional optics hardware in combination with suitable image reconstruction algorithms. We highlight that the seemingly extra information in the computational imaging model is obtained by utilizing information in the bit depth of the sensor pixels. We introduce a data domain information measure that includes computational space-bandwidth product (SBP) of the raw detector data as well as bit depth. Our analysis suggests that the reconstruction algorithms essentially redistribute the information in the bit depth dimension to enable superior imaging capability. The proposed information measure can assess or predict the expected imaging performance independent of the reconstruction algorithm and may provide a generic guideline for designing computational imaging systems.