On the polygonal decomposition in low-level processing for image understanding

The development of knowledge-based digital signal processing (DSP) systems is restrained by several factors. Among them, the deficiency of the low-level knowledge representation scheme is one of the most important ones. The theoretical aspects of the construction of low-level prime components for the knowledge-based DSP systems are presented. Development of the proposed model is proceeded by statistical simulations, which shows that decomposition of an n-dimensional signal should be done using n-dimensional spherical homogeneous prime components. The proposed model significantly reduces redundancy of the input information by an effective and unique transformation, in which array of pixel values is mapped onto symbolic objects of prime components. Such mapping decreases the complexity of high-level processing and simplifies hardware requirements in knowledge-based DSP systems