Image contrast enhancement for outdoor vision is important for smart carauxiliary transport systems. The video frames captured in poor weatherconditions are often characterized by poor visibility. Most image dehazingalgorithms consider to use a hard threshold assumptions or user input toestimate atmospheric light. However, the brightest pixels sometimes are objectssuch as car lights or streetlights, especially for smart car auxiliarytransport systems. Simply using a hard threshold may cause a wrong estimation.In this paper, we propose a single optimized image dehazing method thatestimates atmospheric light efficiently and removes haze through the estimationof a semi-globally adaptive filter. The enhanced images are characterized withlittle noise and good exposure in dark regions. The textures and edges of theprocessed images are also enhanced significantly.
展开▼
