Enhanced image resolution to sub-pixel levels with the nearest neighbor pixel deconvolution (NNPD) method

作者：Yu Wang, Jonathan A. Cain · 年份：2026 · DOI：10.1117/12.3093528 · 研究领域：CCD and CMOS Imaging Sensors、Advanced Image Processing Techniques、Digital Holography and Microscopy

Image enhancement is an important capability for defense and security imaging. An ongoing challenge is to image fine structures in a scene that are smaller than the sensor pixel size and are therefore not distinguished by individual sensor pixels. While High-Dynamic Range (HDR) sensors provide new hardware capability to meet this challenge, both in sensor resolution and bit-depth, algorithmic image enhancement continues to be a welcome complement. The Nearest Neighbor Pixel Deconvolution (NNPD) method for image enhancement was developed in the 2010s and has been applied to several defense applications. At its core, the NNPD regroups an image’s Point Spread Function (PSF) pixels according to the distance between the center of the PSF and the corresponding pixels, applies the Fourier Transform (FT) to the image and PSF, and inverse transforms the ratio to obtain a clearer image. The NNPD is a powerful direct inverse filter image processing method—it can enhance image resolution beyond the diffraction limit with fast processing speed and little impact on memory load. This is critical for missions with short operational timelines, especially when coupled with affordable or Size, Weight, and Power (SWaP)-constrained systems. A new development with the NNPD method, reported for the first time in this paper, is the enhancement of image resolution to the sub-pixel level. We demonstrate the technique against open and publicly available HDR imagery from the James Webb Space Telescope.