Visual sensing and understanding play a fundamental role in building an intelligent vision system. Despite great progress in recent years, it is still extremely challenging for existing machine agents to perceive and describe visual information quickly and accurately in the wild open world, in comparison with their biological counterparts, i.e., the human vision system (HVS). When confronted with complex scene changes and diverse visual content, the HVS can easily evaluate the visual input’s visibility, dynamically adapt the retina for higher perceptual quality, quickly locate all interesting objects, and accurately parse their semantic relationships. These abilities attract us to make use of the neural mechanisms of the HVS.
Numerous research projects have been devoted to visual sensing and understanding. However, the most recent success benefited from a deep neural network, which only mimicked the simplified architecture of HVS and works like a ‘black box’ mapping model. With unpleasant visual artifacts or complex scene changes, existing methods usually suffer from significant performance degradation. The recent developments in neuroscience have enriched our knowledge about HVS and offer great opportunities in building interpretable and trustworthy intelligent vision systems; especially for embedding the neural mechanism of HVS into visual perception modeling, imaging quality enhancement, neural network design, and so on. These have become the frontiers in intelligent vision research.
This Research Topic is to solicit novel and high-quality articles that explore neuroscience-inspired visual sensing and understanding methods, aiming to develop interpretable and trustworthy intelligent vision systems. The topics include, but are not limited to:
• Neuroscience studies of visual perception and recognition mechanism, especially those related to the image/video quality assessment, enhancement, and recognition.
• Neuroscience-inspired image/video compact representation and coding.
• Neuroscience-inspired visual attention models and applications.
• Neuroscience-inspired image/video restoration and enhancement.
• Neuroscience-inspired neural network architecture and learning technologies for image/video understanding including the spiking neural network, continual learning, adversarial learning, and so on.
• Neuroscience-inspired sensing and understanding methods for emerging visual media including the point cloud, light fields, AR/VR, and so on.
Visual sensing and understanding play a fundamental role in building an intelligent vision system. Despite great progress in recent years, it is still extremely challenging for existing machine agents to perceive and describe visual information quickly and accurately in the wild open world, in comparison with their biological counterparts, i.e., the human vision system (HVS). When confronted with complex scene changes and diverse visual content, the HVS can easily evaluate the visual input’s visibility, dynamically adapt the retina for higher perceptual quality, quickly locate all interesting objects, and accurately parse their semantic relationships. These abilities attract us to make use of the neural mechanisms of the HVS.
Numerous research projects have been devoted to visual sensing and understanding. However, the most recent success benefited from a deep neural network, which only mimicked the simplified architecture of HVS and works like a ‘black box’ mapping model. With unpleasant visual artifacts or complex scene changes, existing methods usually suffer from significant performance degradation. The recent developments in neuroscience have enriched our knowledge about HVS and offer great opportunities in building interpretable and trustworthy intelligent vision systems; especially for embedding the neural mechanism of HVS into visual perception modeling, imaging quality enhancement, neural network design, and so on. These have become the frontiers in intelligent vision research.
This Research Topic is to solicit novel and high-quality articles that explore neuroscience-inspired visual sensing and understanding methods, aiming to develop interpretable and trustworthy intelligent vision systems. The topics include, but are not limited to:
• Neuroscience studies of visual perception and recognition mechanism, especially those related to the image/video quality assessment, enhancement, and recognition.
• Neuroscience-inspired image/video compact representation and coding.
• Neuroscience-inspired visual attention models and applications.
• Neuroscience-inspired image/video restoration and enhancement.
• Neuroscience-inspired neural network architecture and learning technologies for image/video understanding including the spiking neural network, continual learning, adversarial learning, and so on.
• Neuroscience-inspired sensing and understanding methods for emerging visual media including the point cloud, light fields, AR/VR, and so on.