Smoothing Noisy Images Without Destroying Predefined Feature Carriers

We address the problem of smoothing gray-level images without destroying feature carriers. Smoothing is performed to suppress high, spatial-frequency noise in the image, whose relevant features contain high spatial-frequency components. The separation is obtained by using a heuristical image-surface geometry criterion over 5x5 mask. Pixel classification results with bit-fields associated with image processing tasks such as noise suppression, edge and/or some 2D-features extraction. We demonstrate the results on standard benchmark image disturbed by uncorrelated gaussian noise. Peformance of some filters applied to feature-less domains of the image is compared.