Camera array-based method and device for capturing high dynamic megapixel video

Abstract

The invention discloses a camera array-based high dynamic one-gigapixel video acquisition method and device. The multi-scale camera array has a reference camera and at least one telephoto camera, wherein the method includes the following steps: at least The telephoto camera image of a telephoto camera is down-sampled; the position of at least one telephoto camera in the reference camera is obtained through a matching algorithm; the image block at the position in the reference camera is up-sampled to obtain a reference block; the telephoto The camera image is aligned with the reference block to embed the telephoto camera image into the reference camera image, and perform image enhancement and color correction to obtain the video capture result. This method can realize video synthesis of one billion pixels, greatly improves the flexibility of the camera array, enables the cameras to be freely allocated to areas in need, increases the utilization rate of CMOS sensors, and has a very high degree of parallelism.

Description

technical field [0001] The invention relates to the fields of computer vision and optical technology, in particular to a camera array-based method and device for capturing high-dynamic, one-gigapixel video. Background technique [0002] The "gigapixel-level video" capture that has emerged in recent years has broken through the limit of human visual information capture, enabling revolutionary multi-scale ultra-high resolution (two orders of magnitude higher than existing cameras) video capture, wide-space capture Both range and high spatial resolution will be satisfied, which can greatly improve the performance of visual computing and promote the technological development of computer vision and artificial intelligence in various fields such as security monitoring, drones, and unmanned vehicles. [0003] The design of camera arrays is a very important problem in the fields of computer vision and optics. The single-lens camera in the related art is limited by the optical syste...

AI Technical Summary

Problems solved by technology

However, this approach mainly has the following problems: (1) poor flexibility: the camera position and angle are fixed

In actual use, it is not possible to allocate and adjust each camera according to actual needs, which may result in a waste of some camera resources; (2) Low camera usage: there needs to be a large enough overlapping area between adjacent cameras (20 %), which greatly reduces the usage rate of the camera CMOS (Complementary Metal Oxide Semiconductor, Complementary Metal Oxide Semiconductor) chip; (3) Color correction is difficult: because the sensor and its parameters (exposure time, white balance, CMOS itself Sensitivity, etc.) are different, so even if adjacent cameras capture the same object, its color and brightness will be different

In addition, images captured by telephoto cameras tend to have darkened edges

Therefore, in order to make the stitching results have a smooth transition, color correction is required, and there is no global reference. Color correction can only rely on the overlapping areas between adjacent cameras, and the correction results are biased, which needs to be resolved

Images