Porosity detection by using Improved Local Binary Patterns Farshad Tajeripour 1 , Shervan Fekri Ershad 2 1. Department of Computer science and electronic engineering, Shiraz University Shiraz, Iran 1. tajeri@shirazu.ac.ir 2. Department of Computer science, engineering and IT, Shiraz University Shiraz, Iran 2. shfekri@shirazu.ac.ir Abstract: - Texture defect detection became one of the problems which has been paid much attention on by image processing scientists since late 90s. Since now many different methods have been proposed to analysis and classification textures. An approach which provides good features to classification is local binary patterns. In this paper an approach is proposed to detection porosity in stones by using the improved form of local binary patterns features. The proposed approach includes two stages. First of all, in train stage, by applying local binary pattern operator on absolutely porosity less images, the basic feature vector is calculated. After that, by image windowing and computing the non-similarity amount between these and basic vector, the porosity-less threshold is computed. Finally, in test stage, by using the porosity-less threshold the porosities is detected on test images. In the result part, the accuracy rate of proposed approach is computed by applying on some captured images and compared with some previous methods. High detection rate, low time complexity, rotate invariant and noise insensitive are advantages of proposed approach. Also, the proposed approach can use for every case of defect detections or visual classification. Keywords: -Defect Detection, Feature extraction, Porosity, Local Binary Pattern, Visual Inspection 1 Introduction Any hole, damage and slot in stone are called porosity. The porosity amount is too important for architectonic stones. Because, the quality of structure is depended to this. Also, the strength of structure or building against earthquake and torrent is depended to porosity amount. The porosity amount is computed by equation (1). Porosity Amount = ୔.୅(୫ మ ) ୗ.୅ (୫ మ ) ∗ 100 (1) Where, P.A means the porosity area which computed in Meter Square measure and S.A is full stone area that computed in same measure. So porosity amount is percental. Some examples of porosity are shown in fig (1). Now, in near all of stonecutting factories, the porosity amount computes by human. So, it’s necessary to propose a visual inspection approach to decrease time and money complexities and increase detection accuracy rate. According to stone images, the porosity is categorized as defect. So, the proposed approach should be a visual defect detection algorithm which names visual inspection algorithm. (a) Fig.1 example of porosity in orange travertine stone Consequently, since now, many different approaches are proposed for defect detection for different cases. For example, Zhaoa and Yeb [1] are proposed an approach for wood defect detection recognition. In [2] the authors offered an accurate method for ceramic tiles visual inspection system. In [3], [4] and [5] some defect detection approaches are proposed for cases such as leather, fabric textile and boiler. In [6] the techniques used to texture analysis and defect detection are discussed in four categories, statistical approaches, Structural approaches, filter based methods, and model based approaches. Table 1 shows a summary list of some of the key texture analysis methods that have been applied to Texture Recent Researches in Communications, Electronics, Signal Processing and Automatic Control ISBN: 978-1-61804-069-5 116