This paper describes the design of small convolution kernels for the r
estoration and reconstruction of Advanced Very High Resolution Radiome
ter (AVHRR) images, The kernels are small enough to be implemented eff
iciently by convolution, yet effectively correct degradations and incr
ease apparent resolution, The kernel derivation is based on a comprehe
nsive, end-to-end system model that accounts for scene statistics, ima
ge acquisition blur, sampling effects, sensor noise, and postfilter re
construction. The design maximizes image fidelity subject to explicit
constraints on the spatial support and resolution of the kernel; The k
ernels can be designed with finer resolution than the image to perform
partial reconstruction for geometric correction and other remapping o
perations, Experiments demonstrate that small kernels yield fidelity c
omparable to optimal unconstrained filters with less computation.