FPGA implementation of an interval type-2 TSK fuzzy neural system for illuminant estimation and white balance

This paper presents FPGA hardware implementation of a type-2 TSK fuzzy neural system for illuminant estimation, i.e., color temperature, based only on the contents of the image along with performing white balance for the image. The implementation is conducted on an FPGA development platform with Xilinx artix-7 XC7A200 chip. Experiments and comparative study are performed. It is shown that the proposed FPGA design can operate in real-time. For generic 1080p image, it can reach the frame rate of 51.598 frames per second at maximal operation frequency for illuminant estimation and white balance. This resolves the slow execution speed of the proposed algorithm on software platform. The results demonstrate major advancements of the proposed approach over previous and existing methods.