Pixel Range Calculation and Pivot Distribution Count Methods for Texture Analysis and Classification of Images

Abstract

Nowadays, Digital image processing techniques have many usages. The digital signals are stored in an organized manner to form an image. The methods interpret those digital signals to extract the features from the digital images. Different approaches are followed to extract features from images to analyze image property, classification of images and medical diagnosis. newlineThis thesis analyzes the images using fractal dimension estimation methods, which can predict the smoothness or roughness of objects or surfaces. The fractal technique has been used as a feature extraction technique, which extracts the feature from images and provides texture signatures for each image in fractal dimension. Those individual texture signatures are further used to map with a set of images to classify them. Pixel Range Calculation (PRC) and Pivot distribution Count (PDC) have been proposed for Texture analysis, classification, and medical diagnosis purposes. The PRC technique has achieved the highest classification accuracy of up to 99.89% while tested for the Brodatz texture dataset. Similarly, the PRC method has gained the most accurate result for other datasets compared to other state-of-art techniques. newline newline