Author(s)

Hanshuang Jia, Ka Zhang, Suiming Yang

Corresponding Author

Hanshuang Jia

Abstract

The enhancement of current digital images is not limited to the single function of highlighting the high-frequency information of the image, it should also suppress noise to a certain extent. It therefore requires a lot of user interaction and is sensitive to noise, such as active contour-based image segmentation methods. This causes great trouble for further applications such as image classification, segmentation, and recognition. This also happens to be a key technology for intelligent interfaces when studying human-computer interaction. In recent years, many edge detection methods based on new theories and technologies have appeared, such as edge detection methods based on multi-resolution analysis and wavelet transform, fuzzy theory, mathematical morphology, artificial neural network and other theories. However, the commonly used multi-scale method has a large amount of calculation, which is to perform linear convolution of the image with a Gaussian function with adjustable width. When performing multi-scale smoothing, it is easy to blur important details of the image. With the development of computer vision technology, image understanding, image recognition, target tracking and other artificial intelligence neighborhood research hotspots. Statistical model can better describe the characteristics of natural images and has attracted extensive attention of researchers. The intensity component of the original image is enhanced in multi-scale, so that the detail information and color fidelity of the image are enhanced; However, the enhancement of dark areas in the image is not obvious, and the image noise is easy to expand. This paper mainly focuses on a main research field of weak illumination image enhancement technology using mathematical morphology for guided filtering.

Keywords

Multi-scale gradient, Guided filtering, Picture