1 Kushagra, Ashish, Hakirat International Journal of Innovations & Advancement in Computer Science IJIACS ISSN 2347 – 8616 Volume 2, Issue 3 March 2014 Texture Classification using Local Binary Pattern with Genetic Algorithmic Approach Kushagra Sharma Ashish Verma Dr. Harkirat Kaur Grewal Dept. of CSE, SSIET, PTU Dept. of CSE & IT, SSIET, PTU Dept. of CSE, SSIET, PTU Dera Bassi, India Dera Bassi, India Dera Bassi, India Kush0466sharma@gmail.com Ashish.ver@gmail.com harkiratgrewal@yahoo.in ABSTRACT Nearest neighbor classifier is used for classification. Local Binary Pattern (LBP) has been widely used in texture classification because of its simplicity and computational efficiency. The nearest- neighbor classifier is the simplest of all algorithms for predicting the class of a test example. The training phase is trivial: simply store every training example, with its label. By coupling the LBP and genetic algorithm for texture analyses and nearest neighbor for texture classification, a more optimized Texture is an important visual clue for vision system. However there is no clear understanding of the nature of texture and texture analysis is still an unsolved problem. As an emerging problem solving method, genetic algorithm has been successfully adapted to various complex tasks in classification and image analysis. texture classification method is presented. Keywords Texture Classification, Local Binary Pattern, Genetic Algorithm, Adaptive LBP INTRODUCTION Texture has been widely used in English to refer to the visual or tactile characteristics of a surface it also can mean a musical pattern, a composite of poetry, and more generally, a basic scheme or structure [14]. In this study, texture refers to the characteristics of the appearance of a surface, which is an important part of visual perception. Generally there are two major approaches to texture analysis: the macro texture approach and the micro texture approach. The macro texture approach is based on the view that texture is the composition of textural primitives. However there are few successful techniques for this approach as it faces at least two complex problems, one of which is the need to identify the textural primitives while the other is the description of the spatial relationship between these primitives. In contrast, the micro texture approach measures textures without identifying textural primitives. The features, for example statistical or structural features, are extracted from the texture images or regions rather than pre-defined texture elements. The micro texture approach has been extensively developed over the last three decades and appears to be much more promising than the macro texture approach. There are an enormous number of texture analysis methods under this category although none predominates. LBP is the method used for feature extraction. This can be done by many methods but each has its own advantages or limitation. Hence we need to use a method which gives more accurate results in less time. So, a method is needed which gives more accurate results. Many methods can be used which we have already studied in the previous chapters. Hence we have to select a feature extractor on the basis of: 1. Accuracy 2. Speed 3. Comprehensibility 4. Time In Summary, the thesis aim to implement the following: LBP extract the features of the input image. After extraction of features, Genetic Algorithm selects the most prominent features and ignores irrelevant and unnecessary features and applies cross-over and mutation operation to find out more optimal features. Hence, the Nearest Neighbor Classifier is used for classification. The above technique gives more accurate results. Traditional LBP codes the sign of the local difference and uses the histogram of the binary code to model the given image. However, the directional statistical information is ignored in LBP. Some directional statistical features, specifically the mean and standard deviation of the local absolute difference are extracted and used to improve the LBP classification efficiency. In addition, the least square estimation is used to adaptively minimize the local difference for more stable directional statistical features, and we call this scheme the adaptive LBP (ALBP). The nearest-neighbor classifier is the simplest of all algorithms for predicting the class of a test example. The training phase is trivial: simply store every training example, with its label. To make a prediction for a test example, first compute its distance to every training example. Then, keep the k closest training examples, (where k<=1). Look for the label that is most common among these examples. This label is the prediction for this test example.