IMAGE-PROCESSING OF PARTIALLY PERIODIC LATTICE IMAGES OF POLYMERS - THE STUDY OF CRYSTAL DEFECTS

High resolution electron microscopy (HREM) is becoming an important technique for the study of crystal defects in polymers. Because of radiation damage, lattice images of polymer crystals suffer from a lack of contrast and a low signal-to-noise ratio that can be improved by digital processing. We investigate the limits and benefits of such methods based on manipulation in Fourier space where periodic features contained in the image can be recognized and separated from the noise. We have calculated the effects of geometrical factors describing the shape and extent of lattice domain(s) present in the image on its discrete Fourier transform by taking into account digitization parameters. Lattice images were obtained from polyparaxylylene (PPX) single crystals which, under the growth conditions we used, contain defects as twin boundaries. Also, a crystal edge and a pair of edge dislocations have been imaged. Image processing methods used were filtration and truncation in Fourier space. The dislocation dipole is well restored by Fourier filtration but whenever the image contains different lattice domains such as rotated parts (twin) or noncrystalline zones, Fourier filtration introduces artefacts of two types: (1) lattice bleeding due to averaging effects results in a loss of crystal/crystal boundary or crystal edge definition, and we show that in these cases better results are obtained by Fourier truncation; (2) a ghost lattice related to the noise components of the image is created by the filter and appears on the reconstructed image at every place where the crystal lattice information is weak or nonexistent. © 1990.