改进暗通道先验的大幅面倾斜影像快速去雾

打开文本图片集

Fast dehazing for large format oblique images based on improved darkchannelprior

HUANG Huang1， OUYANG Huan 1，2* ， DONG Yang²， HE Xin 1 ， ZHAO Xinghao1

（1.93920 Troop，Hanzhong723213，China; 2.Information EngineeringUniversity，Zhengzhou 45ooOl，China） * Corresponding author， E-mail： ouyanghuan@whu. edu. cn

Abstract：The traditional dark channel dehazing algorithm is limited by its inability to adequately remove haze from the distal regions of large-format oblique images and suffers from low processing eficiency. To address these challenges，an improved dark channel prior-based dehazing algorithm for oblique images was proposed in this paper. Initially，a rapid compensation method for the dark channel was employed，utilizing block-level and pixel-level window weighted fusion to enhance transmittance estimation. Subsequently，the dark channel prior was integrated with the radiative transfer equation to establish a haze distribution model and calculate transmittance.The atmospheric light value was then estimated to complete the dehazing process. Furthermore，a parallel computing scheme was designed to enhance dehazing speed. Experimental results indicate that the proposed algorithm effctively mitigates the effects of uneven haze distribution，thus significantly improving the quality of hazy images.In scenes with dense haze，image entropy and average gradient values increase by approximately 22.7% and 30.0% ，respectively.Additionally， the dehazing speed during batch processing is approximately 11 times faster than that of traditional dark channel algorithms， demonstrating a substantial enhancement in processing eficiency.

Key words： image dehazing； large format oblique images；radiation transfer equation；dark channel prior；parallel operation

1引言

倾斜摄影测量是通过在飞行平台上搭载多台传感器，同时从不同角度采集影像，获取完整、详尽、准确的多视角信息，而后通过区域网联合平差、多视匹配、三角网构建、纹理映射等步骤，最终实现实景三维建模。（剩余11301字）