광학적 모델링과 대비향상 기법을 이용한 계층적 야간영상 개선 기법

Abstract

Demands for implementing intelligent unmanned system in military surveillance area, security area using CCTV, vision system area for vehicles, and etc. are increased day by day. In order to satisfy these demands, a lot of researches about the computer vision technology have been accomplished.

The computer vision technology is defined as a technology which recognizes objects by analyzing images acquired from camera and understands the corresponding scenes. One of the problems necessary to comprise an automation system for processing vision information lies in difficulty to analyze the object subject to the influence of low-contrast or irradiation under low-level light circumstance. Therefore, problems about contrast enhancement and irradiation must be solved for improving recognition performance against low light-strength circumstance.

In this paper, a hierarchical processing technique is proposed so as efficiently to improve irradiation and low-contrast characteristic occurred in night vision image. The suggested processing technique is as follows: At first, spawn weight is calculated to solve excessive irradiation problem occurred around light sources of night image. At second, airlight value and transmission value are computed by applying statistical characteristic observed in night image to optical modeling. During this process, regions of texture image and nontexture image are calculated using median filter and applied to the image under processing. A series of processes mentioned above is accomplished with the Gaussian pyramid processing. At last stage, contrast is expanded by applying CLAHE(Contrast Limited Adaptive Histogram Equalization).

Several experiments convinced that the suggested method suppressed irradiation phenomenon occurred in night image and enhanced the performance in both visible and quantitative aspect.