Video synthesis method, device and equipment

• Main Authors: ZHAO MANTANG, LU YONGQUAN, MU KAIHUI
• Format: Patent
• Language: Chinese; English
• Published: 29.08.2023
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; ELECTRIC COMMUNICATION TECHNIQUE; ELECTRICITY; IMAGE DATA PROCESSING OR GENERATION, IN GENERAL; PHYSICS; PICTORIAL COMMUNICATION, e.g. TELEVISION

The invention provides a video synthesis method, device and equipment. The method comprises the following steps: inputting a plurality of frames of images into a pre-constructed target image matting model to obtain a plurality of target objects in different states; wherein the pre-constructed target image matting model comprises a convolution gating circulation unit and an up-sampling unit; the target object is obtained by processing the multiple frames of images by the convolution gating circulation unit and the up-sampling unit; and obtaining a target background image, and fusing the plurality of target objects in different states with the target background image to obtain a target video. Based on the method, video synthesis can be completed in real time with high efficiency, low time delay and high quality. 本说明书提供了一种视频的合成方法、装置及设备。该方法包括：将多帧图像输入至预先构建的目标抠像模型中，得到多个不同状态的目标对象；其中，所述预先构建的目标抠像模型包括卷积门控循环单元、上采样单元；所述目标对象由所述卷积门控循环单元和所述上采样单元处理所述多帧图像得到；获取目标背景图像，将所述多个不同状态的目标对象与所述目标背景图像进行融合，得到目标视频。基于上述方法能够实时、高效、低时延、高质量地完成视