A Spatial Thresholding Method for Image Segmentation
IEEE Transactions on Pattern Analysis and Machine Intelligence
Graphical Models and Image Processing
Image Thresholding by Indicator Kriging
IEEE Transactions on Pattern Analysis and Machine Intelligence
Goal-Directed Evaluation of Binarization Methods
IEEE Transactions on Pattern Analysis and Machine Intelligence
Bilateral Filtering for Gray and Color Images
ICCV '98 Proceedings of the Sixth International Conference on Computer Vision
Least squares quantization in PCM
IEEE Transactions on Information Theory
Hi-index | 0.00 |
We present a new method for image segmentation of X-ray microtomography data that is especially adapted for samples where the feature of interest, in this case the pore size is about the same order of magnitude as the resolution of the data. It combines two filters that are widely used in image processing, Unsharp Mask and Median Filter, to preprocess the tomography data before applying regular Otsu thresholding. For assessing performance, two types of simulated data were constructed, where the true segmentation is known. One data set was purely generic, with tuneable blurring and noise, and a second set used presegmented tomography data taken from chalk samples, where noise was added. Comparison with standard Otsu thresholding and region growing showed superior segmentation results for almost all levels of noise and blurring using the dual filtering approach. It is far more resistant to noise but most important, it retains all the main structural parameters (pore volume, surface area and material volume), even at high noise levels and it provides reliable results across a set of chalk samples with a range of microstructures.