| Summary: | 碩士 === 國立高雄應用科技大學 === 土木工程與防災科技研究所 === 99 === With the advance of the modern spatial information technology, the remotely sensed visible imagery has been used in various filed. The main characteristics of visible imagery are high spatial resolution, stable geometry, and rich information appropriate for human vision. It is also the most important reference for the disaster prevention and management applications. But it is often affected by climate. Among them, especially, fog and haze is the major influence factor when fetching the imagery information. Therefore, effective image dehazing could promote the feasibility of imagery, and decrease the weather condition prerequisite for aerial photography.
There are a lot of methods for image dehazing. In this study, we mainly discuss the Retinex and Dark Channel Prior algorithms which are the most innovative technology in image dehazing. We improved the color correction problems of MSR-based Retinex method corresponding to the human vision. The dark channel prior is also improved based on distribution of fog to promote the calculation speed for huge remotely sensed imagery. This study used various imageries for experiments, and utilized the white balance as an image post-processing tool to make the color of dehazed imagery being appropriate for human visualization. Finally, we applied the image quality assessment index fo objectively evaluate and analysis the image dehazing results.
|
|---|
