A low complexity digital oscillatory neural network for image segmentation

The use of oscillatory neural networks is a recent approach for applications in image segmentation. In this context, the LEGION network is the most useful example. Two attractive aspects are its massively parallel topology and the capacity to separate the segments in time. On the other hand, some limitations that restrict its practical application are found, such as the use of differential equations, implying high complexity for implementation in digital hardware, and limited capacity of segmentation. In this paper, an improved low complexity oscillatory neural network suitable for image segmentation and implementation of digital vision chips is presented. The called ONNIS-GI network offers several advantages, like lower complexity and unlimited capacity of segmentation. Simulations and the implementation of prototypes in FPGA confirm the advantages and the successful operation of the ONNIS-GI network in image segmentation.