Wall-Pasted Cell Segmentation Based on Gabor Filter with Parameter Constraint

Abstract

Gabor filter has been widely used in image processing because of multi-scale, multi-orientation and multi-frequency. To deal with so many parameters, however, especially the frequency parameter, are of great complexity. From the study of one dimensional Gabor filter and its comparison with one dimensional Gauss filter, we proposed a new scheme to simplify its designing work, which mainly deals with the frequency parameter. The filters designed with constrained parameter can approximate the March band effect well. We also applied parameter constraint into 2D Gabor filter designed to reduce the complexity and improve the calculation efficiency. Basing on theoretical analysis and a large amount of experiments, we come to the conclusion that, assisted with rotational filter and Morphological operation, accurate and satisfactory segmentation results can be obtained to the cell images with frequency parameter f taking values between a narrow region. This not only reduces the calculation consumption but also give a direction of how to choose the best segmentation and recognition result