Computer vision tasks are traditionally partitioned into three distinct categories, depending on the types of objects they operate on. It is fairly standard to refer to the computational tasks involving two-dimensional arrays of pixels as low-level tasks. An important low-level task is that of texture discrimination. Our main contribution is to present a parallel implementation of a texture discrimination technique called the Covering-Blanket method on a mesh-based platform, the DAP. The crux of the method is to map estimates of the local fractal dimension of an image lo a feature space that is useful in separating textures. Our results show that the DAP is particularly well suited to handle texture discrimination via the Covering-Blanket method. Our results compare favorably with the best running times of many commercially-available architectures.