CAJ | 학술논문

针对经典的暗原色理论算法在处理雾天图像时色调和亮度失真问题, 提出基于各向异性高斯滤波的暗原色理论雾天彩色图像增强算法. 首先通过容差机制对图像区域进行分割, 根据阈值判断明亮和非明亮区域; 然后引入各向异性高斯滤波, 对透射率图像进行保边平滑处理; 最后再一次引入容差机制, 实现对透射率图像的再次修正,得到准确透射率图像, 进而获得无雾的图像. 主观观察和客观评价结果表明, 在整体和细节方面, 该算法比经典的暗原色算法有更好的处理效果.
침대경전적암원색이론산법재처리무천도상시색조화량도실진문제, 제출기우각향이성고사려파적암원색이론무천채색도상증강산법. 수선통과용차궤제대도상구역진행분할, 근거역치판단명량화비명량구역; 연후인입각향이성고사려파, 대투사솔도상진행보변평활처리; 최후재일차인입용차궤제, 실현대투사솔도상적재차수정,득도준학투사솔도상, 진이획득무무적도상. 주관관찰화객관평개결과표명, 재정체화세절방면, 해산법비경전적암원색산법유경호적처리효과.
To deal with the image hue and brightness distortion problems in the classic dark channel theory algorithm, enhancement dark channel algorithm of color fog image based on the anisotropic Gaussian filter-ing is proposed. Firstly, the image was segmented by the tolerance mechanism and distinguished between bright areas and dark channel areas according to the threshold. Then, the edges of the transmission image were preserved and the others were smoothed by the anisotropic Gaussian filtering. At last, the obtained image was modified to get the accurate transmission image by the tolerance mechanism in order to get a de-fogging image. Through the subjective observation and objective evaluation, the algorithm is better than the classic dark channel algorithm in the overall and details.