Night face video image enhancement and noise reduction method

Abstract

The invention discloses a night face video image enhancement and noise reduction method. Information processing technology field, The method comprises the following steps: firstly, converting a face video image from an RGB space to an HSV space; carrying out BEMD decomposition on the V component of the image; adaptively decomposing into a plurality of IMF components according to a certain rule; filtering the low-frequency IMF component and removing the irradiation component in the low-frequency IMF component; the high-frequency IMF component is textured; The method comprises the following steps: firstly, enhancing information such as details and denoising, reconstructing the processed IMF components to obtain denoised and enhanced V components, performing self-adaptive contrast enhancementon the processed V components, and reconstructing the V components with H and S components subjected to wavelet denoising to obtain denoised and enhanced face video images. According to the method, detail information such as edges and textures can be sharpened while the contrast ratio of the image is effectively enhanced, the definition of the image is effectively improved, and the halo problem in night face image enhancement can be solved.

Description

technical field [0001] The invention belongs to the technical field of information processing, and in particular relates to a nighttime human face video image enhancement and noise reduction method. Background technique [0002] At present, video surveillance technology is widely used in the fields of social governance and public security, and has become an important technical means for public security organs to identify criminal suspects. In the case of good lighting during the day, the collected video images can meet the application requirements, but the quality of the images collected in the evening or at night is seriously deteriorated, and the images show a large number of dark areas and contain noise, resulting in blurred image content and loss of details; Under the artificial light source at night, the collected face images will have high-light areas, making the overall brightness of the image uneven. These problems bring great challenges to the face recognition of ni...

AI Technical Summary

Problems solved by technology

In the case of good lighting during the day, the collected video images can meet the application requirements, but the quality of the images collected in the evening or at night is seriously deteriorated, and the images show a large number of dark areas and contain noise, resulting in blurred image content and loss of details; Under the artificial light source at night, the collected face images will have high-light areas, making the overall brightness of the image uneven. These problems bring great challenges to the face recognition of night video images. Noise reduction methods are important

[0003] At present, image enhancement and noise reduction methods mainly include: enhancement and noise reduction methods based on histograms, enhancement and noise reduction methods based on homomorphic filtering, and enhancement and noise reduction methods based on Retinex theory. These methods have their limitations. If the histogram-based enhancement algorithm does not consider the frequency and details of the image, it is prone to over-enhancement, which will weaken the layering of the image after enhancement; the premise of the application of the enhancement method based on homomorphic filtering is that the illumination is uniform, and the high-light area and dark The image enhancement effect in the area is very poor; the enhancement method based on the Retinex theory is based on the illumination-reflection model. This type of algorithm has halo problems in the edge area of ​​​​the image, and the illumination component is difficult to estimate, which makes the image enhancement effect poor.

Images