Abstract—Generally, an edge-high frequency information in an image-would be filtered or suppressed after image smoothing. This results in noise attenuation, but the image also loses sharpness. This loss of sharpness can impact the usefulness of the processed image for further tasks. Our new anisotropic filtering performs multi-resolution edge analysis and connectivity analysis to make sure that only isolated edge information that represents noise gets filtered out, and the averaging process is within smooth regions not across edges, hence preserving the overall edge structure of the original image. Experimental results obtained from a suite of images but with different signal-noise-ratios (SNR), show that this method is robust to levels of noise and correctly preserves the edges, even for very extremely noisy condition.
Index Terms—Anisotropic filtering, connectivity analysis, multi-resolution, noise reduction.
Bo Jiang is with the Guangzhou Institutes of Biomedicine and Health, Chinese Academy of Sciences, Guangzhou 510530, China (e-mail: firstname.lastname@example.org).
Cite: Bo Jiang, "Anisotropic Filtering with Multi-resolution Edge Analysis and Connectivity Analysis," International Journal of Machine Learning and Computing vol.5, no. 6, pp. 499-506, 2015.