5th Intensive Course on Soil Micromorphology Naples th - 14th September Image Analysis Lecture 9 Grey-Level Morphology and Multi-Spectral Methods

5th Intensive Course on Soil Micromorphology - Naples 2001 Image Analysis - Lecture 9: Grey-Level Morphology Multi-Spectral Methods for Segmentation/Classification useful where X-ray spectra of different colour information (e.g. RED/GREEN/BLUE/ U-V) information is available Part 1 Part 2 Extension of Binary Morphology to Grey-Level Images avoids need to segment images

5th Intensive Course on Soil Micromorphology - Naples 2001 Image Analysis - Lecture 9: Grey-Level Morphology Segmentation (usually by thresholding) and attendant problems Erosion involves stripping pixels from edge of foreground areas according to selected criteria Dilation involves adding pixels to foreground areas Opening involves one cycle of erosion followed by one cycle or dilation roughness aspects of feature are not recovered, no are particles smaller than 2 pixels Closing is the reverse of Opening. Binary Morphology requires

Grey Level Morphology attempts to solve problems of BINARY MORPHOLOGY by removing need for thresholding Grey Level Erosion replaces all intensities within a given mask area by the minimum value in that area Grey Level Dilation replaces all intensities within a given mask area by the maximum value in that area A grey level opening involves an erosion and a dilation phase As with binary morphology, roughness is lost and features tend to become rounded until they finally disappear 5th Intensive Course on Soil Micromorphology - Naples 2001 Image Analysis - Lecture 9: Grey-Level Morphology

Representation of binary morphology for feature sizing 5th Intensive Course on Soil Micromorphology - Naples 2001 Image Analysis - Lecture 9: Grey-Level Morphology

5th Intensive Course on Soil Micromorphology - Naples 2001 Image Analysis - Lecture 9: Grey-Level Morphology Schematic of Intensity Profile along a line

5th Intensive Course on Soil Micromorphology - Naples 2001 Image Analysis - Lecture 9: Grey-Level Morphology Start of Erosion along line

5th Intensive Course on Soil Micromorphology - Naples 2001 Image Analysis - Lecture 9: Grey-Level Morphology Intensity lost after grey-level erosion (blue)

5th Intensive Course on Soil Micromorphology - Naples 2001 Image Analysis - Lecture 9: Grey-Level Morphology Intensity lost after grey-level erosion followed by dilation Blue: Intensity lost: Green: Intensity recovered in dilation

5th Intensive Course on Soil Micromorphology - Naples 2001 Image Analysis - Lecture 9: Grey-Level Morphology Intensity lost after grey-level erosion of diameter 5 Cyan: New Intensity lost

5th Intensive Course on Soil Micromorphology - Naples 2001 Image Analysis - Lecture 9: Grey-Level Morphology Intensity lost after grey-level erosion followed by dilation (diameter 5) Blue/ Cyan: Intensity lost: Green: Intensity recovered in dilation

5th Intensive Course on Soil Micromorphology - Naples 2001 Image Analysis - Lecture 9: Grey-Level Morphology Intensity lost after grey-level erosion of diameter 7 Purple: New Intensity lost

5th Intensive Course on Soil Micromorphology - Naples 2001 Image Analysis - Lecture 9: Grey-Level Morphology Intensity lost after grey-level erosion followed by dilation (diameter 7) Blue/ Cyan/Purple: Intensity lost: Green: Intensity recovered in dilation

Effect of grey-level opening at different radii 5th Intensive Course on Soil Micromorphology - Naples 2001 Image Analysis - Lecture 9: Grey-Level Morphology

a) Radius 9 pixels b) Radius 10 pixels c) Difference Image d) Complete particle loss 5th Intensive Course on Soil Micromorphology - Naples 2001 Image Analysis - Lecture 9: Grey-Level Morphology

Particle size analysis using grey- level morphology th Intensive Course on Soil Micromorphology - Naples 2001 Image Analysis - Lecture 9: Grey-Level Morphology Core sample taken from estuary model. [photograph courtesy of J.Alexander]

5th Intensive Course on Soil Micromorphology - Naples 2001 Image Analysis - Lecture 9: Grey-Level Morphology Halimeda needles from Great Barrier Reef

5th Intensive Course on Soil Micromorphology - Naples 2001 Image Analysis - Lecture 9: Grey-Level Morphology Halimeda needles from Great Barrier Reef

5th Intensive Course on Soil Micromorphology - Naples 2001 Image Analysis - Lecture 9: Grey-Level Morphology Halimeda needles from Great Barrier Reef - partly covered by nanograins

5th Intensive Course on Soil Micromorphology - Naples 2001 Image Analysis - Lecture 9: Grey-Level Morphology Halimeda needles from Great Barrier Reef - partly covered by nanograins

5th Intensive Course on Soil Micromorphology - Naples 2001 Image Analysis - Lecture 9: Grey-Level Morphology Halimeda needles from Great Barrier Reef - fully covered by nanograins

5th Intensive Course on Soil Micromorphology - Naples 2001 Image Analysis - Lecture 9: Grey-Level Morphology Question: Are nanograins biological or chemical in origin? Evidence suggests nanograins increase in size with coverage - hence favouring chemical argument.

5th Intensive Course on Soil Micromorphology - Naples 2001 Image Analysis - Lecture 9: Grey-Level Morphology If particles lost at each radius are stored and finally added (B) - the resulting image should be comparable to original (A). Except: All particles are reduced to their equivalent circular diameter.

Allows alternative methods for segmentation Enables separation of different mineral classes. Can be used in combination with Orientation Analysis as a combination method to overcome problem of large particles 5th Intensive Course on Soil Micromorphology - Naples 2001 Image Analysis - Lecture 9: Multi-Spectral Analysis Requirements: Two or more images of same area at same magnification and pixel resolution and in exact registry. Must be collected with different physical parameters - e.g. wavelength Multi-Spectral Analysis

5th Intensive Course on Soil Micromorphology - Naples 2001 Image Analysis - Lecture 9: Multi-Spectral Analysis Examples: Optical Microscopy: Red / Green / Blue images UV. Electron Microscopy: Secondary Electron Back Scattered Electron Cathodoluminescence X-Ray Maps. Requirements continued:

5th Intensive Course on Soil Micromorphology - Naples 2001 Image Analysis - Lecture 9: Multi-Spectral Methods Require 2 or more different images of same area must be in exact registry e.g. Optical Microscope RED/GREEN/BLUE/UV Or SE / BSE Image and CL or various X - Ray Maps in SEM Multi-Spectral Methods

Hong Kong Marine Clay from M1 unit approximately 1m above upper most palaeo-desiccated layer. BSE Image 5th Intensive Course on Soil Micromorphology - Naples 2001 Image Analysis - Lecture 9: Multi-Spectral Methods

Hong Kong Marine Clay 5th Intensive Course on Soil Micromorphology - Naples 2001 Image Analysis - Lecture 9: Multi-Spectral Methods

5th Intensive Course on Soil Micromorphology - Naples 2001 Image Analysis - Lecture 9: Multi-Spectral Methods BSE ImageX-Ray Maps

5th Intensive Course on Soil Micromorphology - Naples 2001 Image Analysis - Lecture 9: Multi-Spectral Methods From N images and Statistics from M classes Output segmented image may be obtained. Accuracy in segmentation relies on identification of suitable classes, and also sufficient classes

5th Intensive Course on Soil Micromorphology - Naples 2001 Image Analysis - Lecture 9: Multi-Spectral Methods Are these two particles the same material? Classification was set at 98% confidence and some post- processing was done to produce classified image.

Procedure of segmentation is know as Mineral- Segmentation 5th Intensive Course on Soil Micromorphology - Naples 2001 Image Analysis - Lecture 9: Multi-Spectral Methods

5th Intensive Course on Soil Micromorphology - Naples 2001 Image Analysis - Lecture 9: Multi-Spectral Methods

5th Intensive Course on Soil Micromorphology - Naples 2001 Image Analysis - Lecture 9: Multi-Spectral Methods Particle Size Distribution for different mineral species

5th Intensive Course on Soil Micromorphology - Naples 2001 Image Analysis - Lecture 9: Multi-Spectral Methods Binary Mask to assess orientation in matrix outside aggregate. Large mineral grains and voids are black as is aggregate. Binary Mask to assess orientation in matrix inside aggregate. Use Mineral Segmented image to generate binary masks.

Domain Segmentation of Matrix 5th Intensive Course on Soil Micromorphology - Naples 2001 Image Analysis - Lecture 9: Multi-Spectral Methods

Hong Kong Marine Clay a) Matrix orientation c) Quartz grain orientation e) Weighted Quartz grain orientation b) Aggregate orientation d) Feldspar orientation f) Weighted Feldspar orientation 5th Intensive Course on Soil Micromorphology - Naples 2001 Image Analysis - Lecture 9: Multi-Spectral Methods

5th Intensive Course on Soil Micromorphology - Naples 2001 Image Analysis - Lecture 9: Multi-Spectral Methods Index of Anisotropy outside aggregate: inside aggregate: In both cases the predominant orientation is nearly vertical. Vertical direction in field.

5th Intensive Course on Soil Micromorphology - Naples 2001 Image Analysis - Lecture 9: Multi-Spectral Methods When does a particle warrant separate identification from matrix? - depends on pixel resolution/magnification. In supervised classification it is helpful to avoid forced classification as this will identify features / minerals which may have been missed. Some post-processing of image in needed following Mineral- Segmentation to remove noise etc. Concluding Remark on Multi-spectral Analysis.