Ways to improve the effect of defogging

Abstract

The invention discloses a method for improving the defogging effect. The method includes three main processes: dark channel image acquisition, defogging coefficient image acquisition, and ambient light acquisition; and then based on the physical degradation model, through the dark channel image, defogging coefficient image acquisition The image and ambient light are used to obtain a transmittance image, and a dehazed image is obtained through the transmittance image and ambient light; wherein, the dark channel image acquisition process consists of three steps: minimum filtering, maximum filtering, and guided filtering. After minimum filtering on the original image, a sliding window of the same size is used for maximum filtering, and then guided filtering or bilateral filtering is used to obtain a fine dark channel map, which removes the halo phenomenon near the strong edge in the convex area and further suppresses it. The halo phenomenon near strong edges in non-convex areas is solved.

Description

technical field [0001] The invention relates to digital image processing, in particular to an image defogging method. Background technique [0002] There are few current adaptive judgment methods for fog concentration, and the judgment results of existing algorithms are quite different from the results of human subjective judgment. The judgment results are easily affected by noise, which makes the effect of adaptive fog concentration judgment algorithm difficult to meet the needs of automatic defogging. Affects video surveillance image quality. [0003] The existing technology usually extracts a certain feature information from the image, gives the threshold value for distinguishing different fog concentration states through experience, and finally uses the feature information to directly judge the fog concentration type. The judgment result is unstable due to the influence of noise, etc. At the same time, the consistency with human subjective judgment is not good enough, t...

AI Technical Summary

Problems solved by technology

However, there are some problems in this algorithm that seriously affect the image effect: first, only use filters with edge-guided filtering characteristics such as guided filtering / bilateral filtering, and there are still obvious halos at the strong edges of the dark channel image; second , The dark channel principle believes that there are a large number of highly saturated colored objects, shadows and dark areas on the fog-free image. However, this assumption does not hold true for white sky and light-colored buildings. Direct application of this principle will lead to excessive defogging of this part It is not natural enough, and even noise appears; third, the ambient light estimation has a certain influence on the effect of the image restoration by the defogging algorithm. At present, there are few studies on ambient light estimation, and the mainstream algorithm uses the calculation of ambient light with the highest brightness.

[0016] The defogging algorithm based on the physical degradation model, if the estimation of the ambient light is inaccurate, the image will be brighter or darker after defogging

Images