A New Image Quality Index for Objectively Evaluating Despeckling Filtering in SAR Images.

JULIO JACOBO; MARÍA ELENA BUEMI; LUIS GÓMEZ; MARTA MEJAIL

IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING

IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC

Lugar: New York; Año: 2016 vol. 9 p. 1297 - 1307

1939-1404

Synthetic aperture radar (SAR) images are cor-rupted with a multiplicative granular-like noise pattern known asspeckle. The goal for a despeckling filter consists of suppressingthe speckle while preserving all the scene features such as tex-ture, point-type targets, and, especially, edges. There exist severalspeckle filtering techniques and a relevant number of image qual-ity indexes to evaluate the performances of a filtering operationon an SAR image. However, assessing the superiority of a filterover other is not a trivial issue. In this work, we present a newreferenceless estimator (αβ-ratio estimator) based on the ratioedge detector which allows helping in objectively evaluating a fil-ter realization on SAR images. The proposed estimator operateson the ratio image obtained as the point-to-point ratio between theoriginal image (noisy image) and the filtered image. An ideal fil-ter operation on an image implies that, in areas where speckle isfully developed, the ratio image should have the features of purespeckle and no geometric content. The new estimator measuresthe remaining geometric content within the ratio image. This newestimator is easy to compute and it provides an excellent metric torank a filtering operation on real SAR images.