CAJ | 학술논문

暗原色先验去雾方法在户外场景图像去雾方面取得了一定的去雾效果，但该方法在估计近景局部高亮物体透射率时会产生严重误差。本文针对暗原色先验方法的不足首先利用图割理论对粗透射率图进行分块，并使用指导滤波器对细化透射率图进行校正，最后将校正的透射率图带入有雾图像成像模型求得清晰无雾图像。通过将基于暗原色先验的方法、局部对比度最大化的方法和本文中的改进方法进行了实验并对实验结果进行了分析，结果表明本文中的图割分块校正透射率的方法能够更有效地复原因雾霾影响的单幅降质图像。
암원색선험거무방법재호외장경도상거무방면취득료일정적거무효과，단해방법재고계근경국부고량물체투사솔시회산생엄중오차。본문침대암원색선험방법적불족수선이용도할이론대조투사솔도진행분괴，병사용지도려파기대세화투사솔도진행교정，최후장교정적투사솔도대입유무도상성상모형구득청석무무도상。통과장기우암원색선험적방법、국부대비도최대화적방법화본문중적개진방법진행료실험병대실험결과진행료분석，결과표명본문중적도할분괴교정투사솔적방법능구경유효지복원인무매영향적단폭강질도상。
Dark channel prior method achieved a certain effect to dehaze image of outdoor scene, but this method wil produce errors when estimating the transmission of white object located in closing scene. In order to improve the degraded imaging quality which is caused by the influence of fog and haze weather, we introduce an efficient method based on dark channel prior. According to the method of dark channel prior, we segment the transmission map using graph cuts theory and then modify the refined transmission map by guided filter, at last take the modified transmission map into the physical model of atmosphere to get the result of haze free output image. We can see the result from the experiment, it shows that our method could dehaze the degraded image effectively.