A multi-image super-resolution method based on depth information

Abstract

The invention discloses a multi-image super-resolution method based on depth information; the method comprises the following steps: a, obtaining a group of low-resolution images of different reference viewpoints in a same scene and a corresponding group of depth images in a same resolution; b, using depth information to determine parallax sizes of each pixel projected to assigned viewpoint positions in the image of each reference viewpoint, thus obtaining the position of the pixel in the assigned viewpoint; c, projecting the low-resolution image of the reference viewpoint to the assigned viewpoint, and fusing information of all images so as to obtain a high-resolution image. The multi-image super-resolution method can replace an image registering method in a corresponding super-resolution technology in the prior art, is good in super-resolution effect, simple and practical, and has very strong universality.

Description

technical field [0001] The invention relates to the field of image processing, in particular to a multi-picture super-resolution method based on depth information. Background technique [0002] In many occasions, due to the limitation of equipment or working environment, the obtained image quality is very low, which is not conducive to the next step of work, so it is very necessary to obtain higher image quality. [0003] The most direct way to obtain high-quality images is to improve the performance of image acquisition equipment, which requires further reducing the size of the image sensor. However, there is a limit in sensor technology. When the sensor size is small to a certain extent, the image quality is also inevitable. This is because as the size of the sensor decreases, the noise does not decrease, so the proportion of noise in the signal increases. At the same time, the cost of this method is high. Therefore, it is a good choice to use the super-resolution techno...

AI Technical Summary

Problems solved by technology

However, although the optical flow method is reasonable in nature, it is too simple and rough, and the accuracy is not enough, which means that the error will be large, so the final super-resolution result will not be satisfactory; This type of method essentially calculates the motion vector of viewpoint transformation. This method is theoretically more accurate, but there is a problem with this method. It assumes that the motion vector of viewpoint transformation of all pixels in the scene is the same, which means It is assumed that there is no depth difference for all pixels in the scene, which is inconsistent with the actual scene. The actual situation is that objects with smaller depth (closer) in the scene have larger parallax after viewpoint transformation, and objects with larger depth have smaller parallax , which means that the same motion vector cannot be used to describe the viewpoint transformation matrix, the final super-resolution result must be wrong, and the problem of motion estimation also directly affects the quality of the final super-resolution (SR) image restoration

Images