Vis Goes with Heat: A New Way to Combine Day and Night Images

You're trying to see a hidden path, but it's either too bright or too dark. This is where combining visible (vis) and infrared (IR) images comes in. Visible light helps us see details in daylight, while infrared lets us spot heat even in darkness. The trouble? These images are like two different languages. They need to be translated to understand each other, a process called registration.Usually, researchers use fancy tools and metadata to mix these 'languages'. But what if we don't have these tools? That's where "DeepFusion" steps in. It's like a clever translator that doesn't need extra information. First, it zooms in on the right parts of the visible image, looking for thermal information. Then, both images are resized to the same dimensions. Finally, a smart computer model, called a deep neural network (DNN), combines them. We tested "DeepFusion" with popular techniques and datasets. It did really well! Plus, we made a new way to measure how good the mixed images are. This isn't just about making pictures look nice. It's about helping machines understand the world better, even in tough conditions.