Video dyeing method, video compression method, video dyeing system, video compression system and electronic equipment

Abstract

The invention provides a video dyeing method. The video dyeing method comprises the following steps: extracting a multi-layer feature map from gray information of a reference frame and a target frame by using a neural network; performing optical flow prediction and optical flow optimization based on the multi-layer feature map; and transmitting the color in the reference frame to the target frame by using a result obtained by optical flow optimization. According to the method, the accuracy of long-time prediction of the optical flow is improved, the fidelity of the generated color video frame is better, the color is more accurate, and the method has extremely high practical application value. When the method is applied to a video coding task, the effect of reducing the code rate can be achieved on the premise of keeping the reconstruction quality.

Description

technical field [0001] The invention relates to a video coloring scheme based on optical flow optimization, in particular to a video coloring method, a compression method, a system and an electronic device. Background technique [0002] Video coloring is the process of coloring video frames with only grayscale information using reference color information. The reference information for video coloring can be manually specified by the user or extracted from a reference image. Video coloring methods based on training and learning have also emerged with the development of deep learning technology. After a literature search on the prior art, Cheng et al. published "Deep Colorization" at the International Conference on Computer Vision (ICCV) in 2015 and proposed to use deep learning technology to solve the coloring problem. The deep neural network generates color information of grayscale images, and uses joint bilateral filtering to further improve the color information. Zhang ...

AI Technical Summary

Problems solved by technology

[0003] However, among the above methods: the coloring method based on the deep neural network is limited by the data set used for training. If it is applied to a video that is significantly different from the color distribution of the training data set, there will be obvious color difference in the coloring process.

In addition, in the case of long-term prediction based on optical flow, the coloring accuracy drops significantly

Images