High-resolution Large-scale Video Images Enhancement Technology Based On Learning

Image enhancement technology is an important research direction in the field of computer vision.High resolution represents high quality and high definition.High definition video images not only have a wide range of applications in military,medical,surveillance,astronomy,etc.,but also bring a more comfortable visual experience to entertainment life.However,the way to improve image resolution by improving hardware devices(such as developing image sensors with higher resolutions)is not only expensive to develop,but also complicated in manufacturing process.Therefore,how to use the existing camera equipment to achieve ultra-high resolution and large scene video images through cost-effective image enhancement technology has become a hot issue in current research.Considering the importance of image resolution,this paper studies high-resolution video images enhancement technology and proposes two innovative and effective image enhancement techniques,which are learning-based multi-frame image super-resolution technology and many-camera based distributed stitching technology.Based on the mainstream super-resolution algorithms,this paper proposes a multi-frame image super-resolution technology that combines optimal reconstruction with neural network and enhances the high-frequency detail information of the image by introducing spatial prior information.The algorithm achieves excellent reconstruction performance on both subjective visual and objective indicators and is 5dB-7 dB higher in the standard dataset than the mainstream optimal super-resolution algorithm in PSNR.Based on the idea of using multiple inexpensive industrial cameras to collect local scenes and then distributed stitching,this paper proposes a distributed stitching technology based on a many-camera system.By introducing a priori information of spatial locations,a plurality of partial images are merged into a panoramic image of a large scene.The technology achieves high-efficiency and convenient realization of giga-level pixel acquisition that cannot be achieved by ordinary cameras through image fusion and splicing technology.It can be used for real-time live broadcast of large-scale sports events,real-time monitoring of important scenes such as entrance and exit of large stations.In this paper,the above two techniques are tested and analyzed through experiments.Compared with other super--resolution and image stitching methods,this paper has better algorithm performance while ensuring real-time efficiency.