|Proposed Statement of Research
Molly M. Dickens
Texas Tech University
Design of an optimal filter for restoration of degraded images
Real images captured by any image acquisition system are corrupted with noise that can not be perfectly modeled and therefore the restoration of the actual images is considered as an ill-posed problerm Coherent illumination, in particular, results in speckle noise in images acquired by imaging systems with wavelegth limitations for resolving the rnicroscale of the object roughness. Many caherent imaging systems, such as ultrasound medical images, synthetic aperture radar images for remote sensing, astronomical images, and all laser illuminated images,are widely used. Although some researchers have assumed a multiplicative model for speckle noise, it has been shown that such a model is erroneous in many practical situations where the object roughness can not be resolved by the imaging systems. A number of adaptive speckle reduction filters have been developed for intensity speckle images where the phase information is not present. When independent speckle samples are considered, noise smoothing filters have been derived based on the image statistics and assuming a multiplicative model involving a nonstationary mean, nonstationary variance image model.
Recently a different filtering approach based on multiresolution nonlinear mathematical morphology using new connectivity preserving operators used in an alternating sequential filtering ( ASF ) fashion has demonstrated superior performance over commonly used median filters in removing speckle noise from multispectral real sythetic aperture radar (SAR) images . The statistical properties and filtering performances of simple 1- D and 2 D, binary and multilevel, basic morphological operators have also been analyzed and and compared with median filters . We already have preliminary results of the performances of this newly designed connectivity preserving filters, median filters, and linear wavelet filters for removing speckle noise from ultrasound images. The results demonstrate superior performance of the mophological filters over the other two. It is also noteworthy that morphological filters can be developed based on the structural properties of the image without assuming any specific speckle model.
However, statistical properties of the new connectivity preserving filters have not been analyzed yet. I will be involved in analyzing the statistical characteristics of the images and the filters so that an optimal filter for removing speckle noise can be developed. Such filters will have wide use in enhancing the quality of a broad class of images including satellite, remote sensing, and astronomical images.
Wednesday, 26-Mar-2003 21:50:18 CST
CSR/TSGC Team Web