CN-121998860-A - Edge enhancement and denoising method for pixel level of fisheye lens image

CN121998860ACN 121998860 ACN121998860 ACN 121998860ACN-121998860-A

Abstract

The invention relates to the technical field of image processing, in particular to a pixel-level edge enhancement and denoising method of a fisheye lens image, which comprises the steps of acquiring original fisheye image data and preset fisheye lens optical distortion parameters; the method comprises the steps of weight generation, determination of distortion stretching rate of pixel points according to optical distortion parameters, generation of a space density weight graph, threshold judgment and denoising, determination of a target denoising threshold according to the space density weight graph, denoising of original fisheye image data to generate an intermediate denoising image, convolution kernel generation, self-adaptive surface convolution kernel generation according to the optical distortion parameters, edge enhancement processing of the intermediate denoising image by utilizing the self-adaptive surface convolution kernel, and generation of an enhanced image.

Inventors

LIN BIQIANG
YE SUNHUA
FU ZHISEN
ZHENG JIANSONG
LIU QINGBO

Assignees

厦门爱劳德光电有限公司
上饶爱劳德光电有限公司

Dates

Publication Date: 20260508
Application Date: 20260408

Claims (9)

1. The method for enhancing and denoising the edge of the image pixel level of the fisheye lens is characterized by being applied to image processing equipment, and comprises the steps that the image processing equipment acquires original fisheye image data and preset fisheye lens optical distortion parameters; the image processing equipment determines the distortion stretching rate of pixel points in the original fisheye image data according to the optical distortion parameters of the fisheye lens; the image processing equipment generates a space density weight graph according to the distortion stretching ratio, and determines a target denoising threshold according to the space density weight graph; The image processing equipment performs denoising processing on the original fisheye image data according to the target denoising threshold value to generate an intermediate denoising image; the image processing equipment generates a self-adaptive curved surface convolution kernel according to the optical distortion parameters of the fisheye lens; the image processing device performs edge enhancement processing on the intermediate denoising image by utilizing the self-adaptive curved surface convolution check to generate an enhanced image.
2. The method for enhancing and denoising edges at pixel level of an image of a fisheye lens according to claim 1, wherein determining a target denoising threshold according to the spatial density weight map comprises determining whether a weight value in the spatial density weight map is greater than or equal to a preset weight threshold by the image processing device; if yes, the image processing equipment determines that the target denoising threshold value is a preset first denoising threshold value; If not, the image processing equipment determines the target denoising threshold value to be a preset second denoising threshold value, wherein the first denoising threshold value is larger than the second denoising threshold value.
3. The method for enhancing and denoising the edge of the pixel level of the image of the fisheye lens according to claim 1, wherein the generating the self-adaptive curved surface convolution kernel according to the optical distortion parameters of the fisheye lens comprises the steps that the image processing device determines an optical distortion center according to the optical distortion parameters of the fisheye lens; The image processing equipment determines a target deformation direction according to a radial vector in the radial tangential orthogonal coordinate system; And the image processing equipment carries out deformation processing on a preset initial convolution kernel according to the target deformation direction to generate the self-adaptive curved surface convolution kernel.
4. The method for edge enhancement and denoising at the pixel level of a fisheye lens image according to claim 3, wherein said performing edge enhancement processing on said intermediate denoising image using said adaptive surface convolution kernel to generate an enhanced image comprises said image processing device determining a tangential direction according to said radial tangential orthogonal coordinate system; And the image processing equipment carries out edge enhancement processing on the intermediate denoising image along the tangential direction by using the self-adaptive curved surface convolution kernel to generate the enhanced image.
5. The method for enhancing and denoising an edge of a fisheye lens at a pixel level according to claim 1, wherein the method further comprises the steps that the image processing device acquires preset reference image data; the image processing equipment carries out weighting processing on the initial structure similarity according to the space density weight graph to generate distortion weighted structure similarity; And the image processing equipment adjusts preset parameters in the denoising processing or the edge enhancement processing according to the distortion weighted structure similarity.
6. The method for enhancing and denoising an edge of a pixel level of a fisheye lens image according to claim 5, wherein the step of weighting the initial structural similarity according to the spatial density weight map to generate a distortion weighted structural similarity comprises the steps of determining whether a weight value in the spatial density weight map is smaller than a preset center determination threshold by the image processing device; if yes, the image processing equipment gives a first evaluation weight preset for the initial structural similarity; if not, the image processing equipment gives a second evaluation weight preset for the initial structural similarity; The image processing device multiplies the initial structural similarity by the assigned evaluation weight to generate the distortion weighted structural similarity, wherein the first evaluation weight is greater than the second evaluation weight.
7. The method for enhancing and denoising the pixel-level edge of the image of the fisheye lens according to claim 1, wherein the step of determining the distortion stretching ratio of the pixel point in the original fisheye image data according to the fisheye lens optical distortion parameter comprises the steps that the image processing device constructs a physical projection mapping model according to the fisheye lens optical distortion parameter; The image processing equipment inputs coordinates of pixel points in the original fisheye image data into the physical projection mapping model, and calculates an actual physical area corresponding to the pixel points; The image processing device determines the distortion stretching ratio according to the ratio of the actual physical area to the preset standard pixel area.
8. The method of claim 1, further comprising the step of sending the enhanced image to a preset machine vision recognition system by the image processing device so that the machine vision recognition system performs target detection processing according to the enhanced image.
9. The method for edge enhancement and denoising at the pixel level of a fisheye lens image according to claim 8, wherein the machine vision recognition system is deployed in a preset autopilot device; the machine vision recognition system performs target detection processing according to the enhanced image, and comprises the steps that the machine vision recognition system extracts semantic features of an edge area in the enhanced image; the machine vision recognition system generates a target detection result according to the semantic features, and generates control instruction data for indicating the running state of the automatic driving equipment according to the target detection result.

Description

Edge enhancement and denoising method for pixel level of fisheye lens image Technical Field The invention relates to the technical field of image processing, in particular to a pixel-level edge enhancement and denoising method for a fisheye lens image. Background Along with the rapid iteration of systems such as automatic driving, security monitoring and the like, various image processing devices carrying fisheye lenses are on the market, and the nonlinear optical distortion of the fisheye lenses has very high requirements on the precision of image processing; The traditional fisheye image processing is mainly based on the following modes of global unified denoising, fixed square convolution kernel edge enhancement and global unified structure similarity evaluation at present; However, the global unified denoising, the fixed square convolution kernel edge enhancement, the global unified structure similarity evaluation and the like have certain defects, for example, the global unified denoising mode can erase weak high-frequency details with extremely stretched edges as noise to cause the loss of key semantic features, the fixed square convolution kernel edge enhancement has the problem that mathematical processing scale is mismatched with physical stretching scale, the artifact caused by optical dispersion is easily amplified in an edge area, and the global unified structure similarity evaluation mode is difficult to adapt to the edge area with forced stretching of a physical space because global smooth setting unified indexes are excessively pursued. Disclosure of Invention The invention aims to provide a pixel-level edge enhancement and denoising method for a fisheye lens image, which solves the following technical problems: The key semantic features of the edge view field are prevented from being irreversibly erased or reconstruction deviation is generated due to mismatch of the mathematical processing scale and the physical stretching scale, the space perception precision under the extreme view field is greatly improved, and the high-reliability feature preservation is ensured. The aim of the invention can be achieved by the following technical scheme: A method of edge enhancement and denoising at a pixel level of an image of a fisheye lens, the method being applied to an image processing device, the method comprising: the image processing equipment acquires original fisheye image data and preset fisheye lens optical distortion parameters; the image processing equipment determines the distortion stretching rate of pixel points in the original fisheye image data according to the optical distortion parameters of the fisheye lens; the image processing equipment generates a space density weight graph according to the distortion stretching ratio, and determines a target denoising threshold according to the space density weight graph; The image processing equipment performs denoising processing on the original fisheye image data according to the target denoising threshold value to generate an intermediate denoising image; the image processing equipment generates a self-adaptive curved surface convolution kernel according to the optical distortion parameters of the fisheye lens; the image processing device performs edge enhancement processing on the intermediate denoising image by utilizing the self-adaptive curved surface convolution check to generate an enhanced image. Further, determining a target denoising threshold according to the spatial density weight map comprises the steps that the image processing equipment judges whether a weight value in the spatial density weight map is larger than or equal to a preset weight threshold; if yes, the image processing equipment determines that the target denoising threshold value is a preset first denoising threshold value; If not, the image processing equipment determines the target denoising threshold value to be a preset second denoising threshold value, wherein the first denoising threshold value is larger than the second denoising threshold value. Further, the self-adaptive curved surface convolution kernel is generated according to the optical distortion parameters of the fisheye lens, and the self-adaptive curved surface convolution kernel comprises the image processing equipment, the image processing equipment and a self-adaptive curved surface convolution kernel, wherein the image processing equipment determines an optical distortion center according to the optical distortion parameters of the fisheye lens; The image processing equipment determines a target deformation direction according to a radial vector in the radial tangential orthogonal coordinate system; And the image processing equipment carries out deformation processing on a preset initial convolution kernel according to the target deformation direction to generate the self-adaptive curved surface convolution kernel. Further, the intermediate denoising image is checked by the self-adaptive surface