Scaling properties of binary and greyscale images in the context of X-ray soil tomography
| Autor: | Juan Carlos Losada, Iván González Torre, Pilar López, Juan J. Martín-Sotoca, Ana M. Tarquis |
|---|---|
| Jazyk: | angličtina |
| Rok vydání: | 2020 |
| Předmět: |
Ground truth
business.industry Binary scaling Soil Science Pattern recognition Context (language use) Rayos X 04 agricultural and veterinary sciences Multifractal system 010501 environmental sciences 01 natural sciences Grayscale Lacunarity 040103 agronomy & agriculture 0401 agriculture forestry and fisheries Segmentation Artificial intelligence Tomography business Ciencias de la tierra Hongo 0105 earth and related environmental sciences Mathematics |
| Popis: | Characterization of the complex soil structure is one the cornerstones of soil science and pore space detection is a crucial step in this process. Synthetic soil image construction has been proved to be an efficient resource for validating different binarization methods given that, unlike in real world, ground truth information is known. In this work, we introduce an improved Truncated Multifractal Method (TMM), to better simulate synthetic computed tomography (CT) soil images and then we generate 150 synthetic images with three different porosities (7%, 12% and 17%), both in greyscale and in binary scale (pore spaces). Synthetic images are then compared with two sets of 260 slides of real CT soil images, in order to validate the goodness of the method. All images are subjected to multifractal analysis where we show a detailed comparative analysis of parameters such as lacunarity, characteristic length and multifractal spectrum, that are calculated both for the whole set of synthetic (greyscale and binary) and for the sets of real CT soil images. With respect to lacunarity, a not previously reported inverse relationship between binary and grey lacunarity is found for this range of porosities. Moreover, we have also reported a new relationship between lacunarity and characteristic length. Similar multifractal results, that we obtain for real CT and synthetic soil images, prove that TMM is a reasonable solution to create simulated CT soil images. Finally, a segmentation test was carried out, using TMM synthetic greyscale soil images and its binary counterpart as ground-truth information, evaluating global (Otsu) and local (Combining Singularity-CA) binarization methods, where we report better performance for the last. Programa Propio (Universidad Politecnica de Madrid) Ministerio de Economia, Industria y Competitividad, Gobierno de España (FIS2017-84151-P) Comunidad de Madrid 6.114 JCR (2020) Q1, 3/37 Soil Science 1.846 SJR (2020) Q1, 6/140 Soil Science No data IDR 2020 UEM |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|