Fisheye image processing method, computer device and computer readable storage medium

Abstract

The invention discloses a fisheye image processing method, a computer device and a computer readable storage medium. The method comprises steps: the region of a next region of interest image and the region of the current region of interest image are determined, whether an overlapped region exists in the next region of interest image and the current region of interest image is judged, and if the overlapped region is determined to exist, data of the overlapped region are stored to a distortion correction data source memory from a backup memory; whether an image beyond the overlapped region is included in the next region of interest image is judged, and if yes, the data of the image beyond the overlapped region are stored to the backup memory and the distortion correction data source memory; and image data stored in the distortion correction data source memory are corrected. The invention also provides a computer device for realizing the above method and a computer readable storage medium. Thus, the fisheye image processing circuit complexity degree can be reduced, the circuit area is reduced, and the image processing efficiency is improved.

Description

technical field [0001] The present invention relates to the field of image processing, in particular to a method for processing images captured by a camera equipped with a fisheye lens, a computer device for realizing the method, and a computer-readable storage medium. Background technique [0002] With the development of photographing equipment, people widely use various wide-angle lenses to take pictures, among which fisheye lens is a wide-angle lens with an extremely wide viewing angle. Photos taken with fisheye lenses are usually called fisheye images. Due to the large field of view of fisheye images, a single fisheye camera can obtain the content of 180-degree field of view information. This feature makes fisheye lenses more and more popular. It is used in autonomous driving, panoramic monitoring, virtual reality and other fields. [0003] However, due to the large distortion of the fisheye image itself, it is necessary to correct the fisheye image through an appropria...

AI Technical Summary

Problems solved by technology

Due to the uncertainty of the mapping relationship and the requirements of the distortion correction algorithm for the data source, when correcting the fisheye image, most of the current processing methods are to cache multiple rows of original pixels or directly obtain the corresponding pixels from the entire image. Distortion correction operation is performed, but this method leads to large image processing circuit area and low read and write efficiency of DRAM

[0004] Specifically, most of the current processing of fisheye images uses caching of multiple rows of original pixels or directly obtains corresponding pixels from the entire image to perform distortion correction operations. For the multi-row original data caching scheme, enough rows of image data must be stored to To ensure that the correct data can be obtained for calculation during the distortion correction process, this requires a very large area of ​​the static memory SRAM to meet the requirements of data storage, and the static memory SRAM needs to be implemented in a rolling manner or by repeatedly loading part of the data by row Data update, complicated operation, introducing too much bandwidth of invalid DRAM

In addition, the depth of the static memory SRAM used in this method is restricted by the image specification, which is not convenient for upgrading the chip specification

However, the method of directly obtaining the corresponding pixels from the entire image to perform distortion correction calculations has large and discontinuous address jumps when acquiring data, low efficiency of DRAM, and also introduces too much invalid DRAM bandwidth.

[0005] In the current way of processing fisheye images, it is better to divide the obtained fisheye images into blocks. A major bottleneck in the block processing of fisheye images is the excessive input bandwidth of the aforementioned DRAM. , is not conducive to the hardware implementation of the fisheye image processing circuit, especially not conducive to the integration of the hardware of the fisheye image processing circuit on the system-on-chip (SOC)

Images