Dynamic Wiener filters for small-target radiometric restoration

Abstract

Small-target radiometric restoration (STRR) seeks to correct imagery for the blurring effects of the sensor and allow more accurate radiometric values to be extracted. This paper describes an STRR approach suitable for imaging systems whose point-spread functions are known and slowly varying across the image. The approach features a dynamic Wiener filter based on the physical properties of the target and its background. The Wiener filter is constructed based on three inputs: the modulation transfer function (MTF) of the sensor, the estimated target spectrum, and the estimated noise spectrum of the target. The target spectrum is approximated using an estimate of the target-background contrast from the detected image and the geometry of the pixels that approximately define the target. The noise spectrum is obtained from the target radiance in the detected image and the relationship between noise variance and image radiance for the sensor.