A GPU-based Fisheye Correction Method

Abstract

The invention provides a fisheye correction method based on GPU, which includes the steps as follows: establishment of a fisheye model; acquisition of a fisheye correction algorithm; image acquisition; UI control; determination of a fisheye correction range; and virtual PTZ. With a GPU as hardware support and based on OpenGL and D3D database development, a fisheye image is loaded as texture to the GPU, and a 3D model of fisheye is established. Corresponding correction formulas are derived for different correction modes. The correction formulas are implemented in the shader of the GPU. Correction operation is carried out using the shader of the GPU to get a result after correction. Compared with the traditional CPU computation, a shared memory method is used when fisheye image data is loaded as texture to the GPU, which can greatly improve the transmission efficiency of the fisheye image from CPU to GPU; and through fisheye correction in the GPU shader, a real-time processing effect can be achieved for fisheye correction.

Description

technical field [0001] The invention relates to the technical field of video monitoring, in particular to a GPU-based fisheye correction method. Background technique [0002] Fisheye lens is an ultra-wide-angle lens that simulates the effect of underwater fish looking up at the water surface. Its typical field of view is 180°, and some exceed 180°, greater than 220°, and even reach 270°. Fisheye lens has many advantages: it has a large viewing angle, can accommodate many scenes, and can adapt to shooting in a small space. Therefore, it is widely used in many computer vision fields such as virtual reality technology, robot navigation, and visual monitoring. Structurally speaking, the fisheye lens is indeed similar to the fisheye structure, and there are two types of fisheye lenses, one is called Circular Image Fisheye (circular fisheye), and the other is called FullFrame Fisheye ( Full-frame fisheye), the viewing angle of its design is basically 180°. Due to the ultra-wide ...

AI Technical Summary

Problems solved by technology

[0003] In the field of security monitoring, homes, public places, transportation hubs, shopping malls and entertainment and other occasions need to use panoramic fisheye cameras to achieve 360°×180° large-scale high-quality panoramic real-time monitoring without dead angles, usually using a single fisheye camera It can replace multiple ordinary cameras with different installation angles. The advantage of the fisheye camera is that the field of view is large, but due to its large field of view, it causes serious distortion and does not conform to the observation habits of the human eye, so the fisheye correction technology becomes Especially important, the traditional fisheye correction technology mostly uses CPU for correction, but there are many defects in using CPU for correction: on the one hand, it is limited by the memory calculation method of the CPU, and cannot correct larger images such as 4k images; on the other hand, it cannot Guarantee the real-time performance of image correction, and there are problems such as high CPU performance consumption, single correction mode, and difficult observation of various parts during the correction process

However, this method has a single correction mode and cannot observe each part

Images