A new heuristic algorithm for porosity segmentation for the colored petro-graphic images is proposed. The proposed algorithm automatically detects the porosities that represent the presence of oil, gas, or even water in the analyzed thin section rock segment based on the colour of the porosity area filled with dies in the analyzed sample. For the purpose of the oil exploration, the thin section fragments are died in order to emphasize the porosities that are analyzed under the microscope. The percentage of the porosity is directly proportional to the probability of the oil, gas, or even water presence in the area where the drilling is performed (i.e. the increased porosity indicates the higher probability of oil existence in the region). The proposed automatic algorithm shows better results than the existing K-means segmentation method.
Top1. Introduction
Geological and petrophysical domains are becoming more appreciative of the use of digital image processing in many of their applications. One example is the extraction of porosity and different components of solid fraction information (grains, matrix, and cement) from the microscopic digital images of their thin sections (Hatfield, 2001). This appreciation came as a trivial result of the increased demand in oil production, especially from non-conventional reservoirs, e.g., within carbonate structures, which requires more in-depth analysis of logs, seismic, and cores where thin sections can be obtained.
Some petrologists can easily estimate macro porosities using microscope images since the porous media possess a very distinct color such as the light blue and pink as shown in Figure 1. However, the process is time consuming and may have a certain degree of error in the final results (Hatfield, 2001).
Figure 1. Samples of colour microscopic images of rock segments
Moreover, generally quite a large number of thin sections are made for every cored well. Therefore, an automatic and efficient technique is required for extracting and quantifying different components of rocks from their thin sections. Hence, this requires a successful technique for dividing petrographic images of thin sections into different regions where each region is nearly homogenous and the union of any two regions is not. This division process is known as digital image segmentation (Gonzalez, 2002), (Bovik, 2005).
Such application of digital image segmentation for carbonate thin section images is getting more attention from researchers.
The proposed method for automatic porosity detection and segmentation was tested on four different real world color microscope images. The color microscopic images with died porosities in blue and one with died porosities in pink are shown in Figure 1. The Figure represents four different thin section images of different rock types, namely, grainstone, dolomitic grainstone, and dolostone.
The colour images belong to the Red-Green-Blue (RGB) domain. Samples 1-3 have resolution of 1550 x 2087 x 3 in JPEG format and Sample 4 has resolution of 774 x 1183 x 3 in JPEG format.
Samples of microscopic images contain different level of porosity which indicates the presence of oil, gas or even water in the particular area.
The proposed novel heuristic algorithm is described and the experimental results show a promising advance of the addressed method in comparison to the known k-means technique.
Top2. Color Based Porosity Segmentation Algorithm Combined With Wavelet De-Noising Analysis
The research leading to this study has been focused on implementing a robust algorithm that is capable of porosity detection in petro-graphic color images. The proposed novel automatic algorithm combines a de-noising method realized with wavelets analysis and the porosity detection based on color analysis. The color based heuristical model is derived taking advantage of the color of the dyed targeted porosities. Instead of developing a complex mathematical model and algorithm that would be detecting the unpredictable contours of the porosities, here the color is used as the parameter for the porosities detection and recognition providing a simpler and efficient technique.