Determining the Direction of Images Using Sum and Difference Histograms and Texture Descriptors

One of the most used statistical approaches in the literature for computing texture information is the Gray Level Cooccurrence Matrix (GLCM). The use of GLCM and textures descriptors calculated from there, allows, among other applications, the detection of the main direction of the elements of texture. This is possible essentially because the GLCM uses different angles to compute the relationship between the gray tones in the image. An alternative to computing this relationship is the Sum of Difference Histograms (SDH). In contrast to GLCM, which is a bidimensional structure, SDHs are one-dimensional. Once there is a relationship between the data stored in both structures, the behavior of the descriptors is similar, and the SDH have a lower computational cost being, for this, more efficient in a number of applications. In this sense, this work leads to a series of investigations regarding the correspondence between these two approaches. Moreover, it investigates which texture descriptors, calculated from the SDH, can collaborate in the process of identifying the predominant direction of a texture.