Space and time efficient data structures in texture feature extraction

Texture feature extraction is a fundamental stage in texture image analysis. Therefore, the reduction of its computational time and storage requirements is an important objective. The Spatial Grey Level Dependence Method (SGLDM) is one of the most important statistical texture analysis methods, especially in medical image processing. Co-occurrence matrices are employed for its implementation. However, they are inefficient in terms of computational time and memory space requirements, due to their dependency on the number of grey levels in the entire image (grey level range). Since texture is usually measured in a small image region, a large amount of memory space is wasted while the computational time of the texture feature extraction operations is unnecessarily raised. Their inefficiency puts up barriers to the wider utilisation of the SGLDM in a real application environment, such as a clinical environment. In this thesis, three novel approaches which are based on dynamic binary trees to organise the textural information extracted from an image region by the SGLDM, are presented and evaluated both theoretically and through their application to the analysis of natural textures and medical images of various modalities. Their novelty is based on their ability to eliminate the redundant information stored in the co-occurrence matrix, due to their dependency only on the number of distinct grey levels in the analysed local region. These are shown to provide efficiency in both storage requirements and computational time compared to the co-occurrence matrix, especially in the analysis of images employing their full dynamic range. Two experiments involving the classification of clinical and non-clinical image data are performed, which clearly show the benefit of analysing images using SGLDM, without reducing their grey level range