Depth image super-resolution processing method based on deep learning

Abstract

The invention belongs to the field of computer image processing, and aims to provide a method capable of carrying out effective super-resolution processing on a depth image by using a corresponding high-resolution color image to obtain sharp depth image edges. Therefor, the adopted technical solution of the invention is a depth image super-resolution processing method based on deep learning. The method comprises the following steps: 1) selecting a certain number of rich-texture depth images and color images corresponding thereto from a public data set; 2) carrying out data enhancement; 3) carrying out data preprocessing on obtained depth and color image pairs; 4) designing a deep convolutional neural network structure; and 5) using the preprocessed data set to train the designed convolutional neural network, inputting the low-resolution depth image and the corresponding color image into the trained network after the convolutional neural network is trained, and outputting a depth imageafter super-resolution processing is completed at an output layer. The method is mainly applied to image processing.

Description

technical field [0001] The invention belongs to the field of computer image processing, and in particular relates to a depth map super-resolution method based on a convolutional neural network using depth color image pairs. Background technique [0002] Depth information is an important information for the visual perception of 3D objects. Applications in autonomous driving, human-computer interaction, 3D scene reconstruction, and virtual reality all rely on high-performance and high-quality depth maps. However, the current consumer-level depth cameras, including Microsoft's Kinect, Asus' Xtion Pro and other TOF cameras based on time of flight (TOF) technology, due to limitations in hardware conditions, the resolution of the obtained depth images is low. Very low, well below the resolution of a color image. This brings great limitations in practical applications, therefore, in order to utilize the depth information data more effectively, the super-resolution method of the de...

AI Technical Summary

Problems solved by technology

In the process of obtaining high-resolution depth maps by the above methods, some methods have the problem of high computational complexity and long calculation time, and some methods also have artificial traces in the results in some cases, and the effect is not ideal. The problem

Images