(1) Marija Delić, Local binary pattern descriptors for texture classification
Local binary pattern (LBP) descriptors are commonly used in texture classification in recent years. They were introduced as descriptors of local image texture and their histograms were proven to be very important texture features. The texture of objects is one of their most important properties. Object recognition and classification relying on texture-based descriptors are used in many applications, such as biomedical image analysis, document image analysis, industrial surface inspection, texture synthesis for computer graphic and animation, etc. A number of methods and techniques for texture analysis were proposed, but many of them are not capable to perform well with real textures or are very computationally complex and take a lot of time for execution. Defining a good texture descriptor is not an easy problem. Real textures can occur at arbitrary spatial resolutions and rotations and varying conditions in illumination. This problem inspired collection of studies which work to find invariance with respect to one or more properties such as spatial scale, orientation and gray-scale. Defining a powerful texture descriptor which can be extracted and classified with low computational complexity is therefore a demanding but challenging task. We are primarily interested in the family of local binary pattern (LBP) descriptors and texture classification. LBPs have been shown as family of descriptors which are suitable for many applications, being computationally attractive and simple. Motivated by the main idea of LBPs and fuzzy set theory, we propose a novel member of LBP family which utilizes and combines these two mentioned independent concepts.