A colorization method of low-light image based on multi-dimensional data association rules

Abstract

The invention discloses a low-light image colorization method based on multi-dimensional data association rules. The steps of the method are as follows: Mining multi-attribute rules of low-light images: firstly, using the sample library composed of different objects in similar scenes, the low-light images are classified based on SVM, and the Apriori optimization algorithm constrained by multi-attribute rules is used to mine the reference color images. A strong association rule set between brightness and color, and introduce category label elements to finally generate a brightness-category-color strong association rule set; colorization of low-light images based on rule mapping: extract the brightness and category of pixels in low-light images , based on the strong association rule set generated by mining, map and generate the R, G, and B color components corresponding to each pixel to generate a color map. The method of the invention can obtain a color image with good colorization effect and high color restoration degree, and the method is easy for hardware conversion, and can realize real-time colorization effect.

Description

technical field [0001] The invention belongs to the technical field of night vision image colorization, in particular to a low-light image colorization method based on multidimensional data association rules. Background technique [0002] Low-light image colorization has always been a research hotspot in night vision technology. The low-light image is the result of photoelectric device imaging, usually a grayscale image. Combined with the characteristics of higher resolution and sensitivity of the human eye to color images, the colorization of low-light images can improve people's cognition of target and scene information , both in the military and civilian fields are of great significance. [0003] At present, the methods for colorizing grayscale images mainly include color diffusion methods based on manual strokes and color transfer methods based on reference color images. Based on the assumption that "if spatially adjacent pixels have similar brightness, their colors wi...

AI Technical Summary

Problems solved by technology

The color diffusion method based on manual strokes requires manual input from the user, and the user's manual input is very demanding. Generally, it needs repeated operations many times to obtain the ideal colorization effect; the color transfer method based on the reference color image is complex and time-consuming. More, and it is difficult to get the ideal colorization effect

Images