现有雾天图像增强的Retinex算法采用固定滤波器，无法适应多种景深和雾化程度的情况。对此，本文提出一种基于暗原色先验模型的Retinex算法。暗原色先验模型反映了雾天图像中雾的分布与景深信息。受此启发，根据局部区域暗原色值设计一种尺度可变滤波器，针对不同景深和雾化区域采用不同尺度的滤波器估算雾天图像的照度分量，实现对雾天图像的增强。分别使用主观观察和客观数据分析方法，将本文算法与HE算法、固定尺度MSR算法进行对比，本文算法在细节增强以及图像整体效果上均优于HE算法和固定尺度MSR算法。%Current Retinex algorithm applied in foggy image enhancement with fixed filter can ’t adapt to the situation of var-ious depth of field and atomization .This paper presents a self-adaptive filter Retinex algorithm based on dark channel prior model . The model of dark channel prior reflects the information of field depth and distribution of atmosphere in foggy image .We are in-spired to design an self-adaptive filter according to the local value of dark channel using different filters in different depth of field and foggy area to estimate the illumination component of the image ,and to achieve the clarity of foggy image .Finally ,we compare the result of the proposed algorithm with the result of HE algorithm and result of fixed filter MSR algorithm using the subjective ob-servation and objective data analysis method .The comparison shows that the result of Retinex algorithm based on dark channel prior has better detail of image and global effect than that of the HE algorithm and fixed filter MSR algorithm .
展开▼
