KEENSENSE
There is a huge market for image enhancement, covering AGV, drone, surveillance, robotic, Metaverse, etc., with the estimated camera number in 2025 will be 500 million, with a compound annual growth rate larger than 19%. The fusion of multiband imagery has become important for navigation, target tracking, object detection, and face recognition because multiband imagery can provide more information than single-band imagery. But the quality of the fused image is not well addressed at this stage. To tackle the current technical difficulties, we are going to develop a high-quality and hardware-friendly multiple-modality image enhancement algorithm to provide better fusion images to meet the requirements of different applications, in which, the data from IR sensors and RGB sensors will be used as inputs, AI-based algorithms will be used to extract multi-scale salient features from the neighboring frames and neighboring pixels for multiband image enhancement, and we will also do algorithm-hardware joint optimization to make the algorithms feasible for hardware implementation. We will provide an optimized embedded GPU-based solution for surveillance, navigation, and AGV applications, and provide proven hardware feasibility to attract SoC and ISP IC manufacturers to improve their products' competition.