Application Of Local Binary Pattern And Human Visual Fibonacci Texture Features For Classification Different Medical Images

Textural information plays a critical role in performing and understanding the analysis for different types of microscopic images. The local binary patterns (LBP) have emerged among the most efficient texture features because of its easy implementation, rotation invariance, and robustness to monotonic illumination changes. However, the LBP is sensitive to noise and nonmonotonic illumination changes, it is unable to capture macrostructural information, and has large feature vector size. The goal of this paper is to (a) present an extended variant of the LBP, called the Fibonacci -p pattern and (b) analyze the LBP and Fibonacci -p pattern based texture features for different medical images such as histopathology images, MRI images, CT images, and mammograms. The performance of the classification system of 251 prostate histopathology is analyzed using evaluation parameters such as accuracy, sensitivity, specificity, positive predictive value, and negative predictive value. On computer simulation, n, an increase in cancer classification accuracy is observed from 87.42% (LBP features) to 96.69% (Fibonacci -p pattern features) while maintaining the computational efficiency. Finally, on comparing with the traditional LBP, the Fibonacci -p patterns have approximately the same computational cost, lesser feature size, and the Human Visual Fibonacci System has robustness to illumination changes, additional texture information, and enhanced edge information.