Yolanda Prieto

Locally adaptive orientation Wiener image filter with local noise estimate

The restoration of images degraded by additive noise has been addressed previously by several authors. In this work, to overcome the consequences that arise from cascading filters sequentially (namely, the fact that the noise behavior changes more and more the deeper into the cascade), while still preserving edges and maintaining low computational requirements, we propose a modified locally adaptive orientation Wiener filter (MAOW). Different masks are applied at each pixel to a local region and the mask yielding a minimum variance is the selected one; thus mask orientations can vary locally. In addition to improve the operation of the AOW, we need a noise estimation that is locally varying. This is obtained by using the information that is already available to us from the quantizer prior to filtering.