Linear filtering of spatially invariant image sequences for feature separation under three types of image noise
R. Olmstead, James B. Farison · 2003
Many important imaging applications in medical imaging, remote sensing and other areas result in a set of images of the same scene with no relative scene-sensor motion and in which the pixel intensities are the linear sum of the contributions of the distinct features in the scene. Such image sets are called linearly additive, spatially invariant (LA SI) image sequences. Previous research has shown, both mathematically and with examples, that a K-image sequence with M distinct features (M<K) can be linearly filtered to extract the individual features from the original image sequence. In those studies, the noise was modeled as Gaussian (uniform over the image scene) or Poisson (dependent on the feature distribution). This paper explores the effectiveness of the same technique for LA SI image sequences with salt and pepper noise, which is neither uniform nor feature dependent, and compares the results with those for the same image scene with equal Gaussian and Poisson noise. It is shown here that the method extracts the features in this case with similar effectiveness.