The Study Of Fog Images Simulation&Haze And Thin Cloud Removal Methods

Fog is an important factor affecting image quality. The images obtained in foggyweather have serious degradation problems. While in such applications as AircraftGuidance need lots of different foggy weather images to simulate the flying andemergency handling processes. The images acquired by remote sensing are easy to beblurred due to the cloud cover. Therefore the study of fog images simulation andrecovery as well as cloud removal is of extremely siginificance.The main contributions of this thesis on fog images simulation as well as haze andthin cloud removal are as follows:1) Realized the fog images simulation using Atomospheric Modulation Transfer(MTF) function. Got hazy, misty and foggy simulated image respectively throughMTF of corresponding weather condition as its degradation model.2) Realized single haze image recovery using Dark Channel Prior algorithm, andsolved problems like having to consume large amounts of storages and time aswell as inaccurate result of the sky region of this algorithm. Acquired moreaccurate Atmospheric value by identifying sky region or the most concentratedhaze area first. Obtained dark channel images using different template processingon the edge area and non-edge area respectively, and optimized the transmissionmap according to the relationship between dark channel image s and depth, whichenormously reduced the computation time. Solved the discontinuous atmosphericimages dehazing problems effectively by processing the glare region andnon-glare region respectively.3) Realized thin cloud removal of remote sensing images using Homomorphic filterand Dark Channel Prior algorithm, and solved the uneven cloud removalproblems by combining the improved dark channel prior with large-scaleGaussian filter method.