Video space-time super-division model construction method, device and equipment and readable storage medium

The invention relates to a video space-time super-division model construction method, a video space-time super-division model construction device, video space-time super-division model construction equipment and a readable storage medium. Then mutual learning of time video super-division and space video super-division tasks is carried out through the upward projection module and the downward projection module, time information and space information are promoted to be completely utilized, and cross-space errors are eliminated; and meanwhile, the image reconstruction module reconstructs a plurality of low-resolution feature maps and high-resolution feature maps which are subjected to mutual iterative learning to generate final high-resolution video frames and low-resolution video frames, so that the reconstruction of a real image is realized, and the efficiency of video space-time super-division is improved. 本申请涉及一种视频时空超分模型构建方法、装置、设备及可读存储介质，由于本申请中先基于特征时间插值网络生成中间帧的初始低分辨率特征图，然后再通过向上投影模块和向下投影模块进行时间视频超分和空间视频超分的任务的相