单景空照影像除雾处理之研究TheStudyofSingleAerialImage.PDFVIP

Conference on Computer Applications in Civil and Hydraulic Engineering 2011 (CCACHE 2011) September 01-02, 2011, Kaohsiung, Taiwan 單景空照影像除霧處理之研究 1 2 3 4 李彥玲 、謝嘉聲、李良輝、張庭榮 摘要 光學遙測影像易受到霧、霾的影響而降質。因此，對影像進行除霧處理，可增加光 學遙測影像資料的有效性和可用性，且可降低氣象條件對航空攝影的限制。影像的除霧 方法有許多種，本研究主要探討Retinex演算法及暗通道先驗法兩種較新的除霧方法。 Retinex理論中以其中的 MSR 方法為基礎，本文改進其影像色偏問題，使除霧影像色彩 更符合人眼視覺效果。暗通道先驗法則針對遙測影像的霧分佈特性使用改善計算速度的 演算法，以利大型影像之計算。以各種不同影像進行影像除霧實驗，並使用影像品質評 估指標對除霧的結果進行客觀的分析與比較。 關鍵字 暗原色先驗；霧成像模型；除霧；圖像清晰度。: The Study of Single Aerial Image Dehazing 1 2 3 4 Lee Yen-Ling , Hsieh Chia-Sheng , Lee Liang-Hwei and Chang Ting-Rong ABSTRACT The visible remotely sensed image is often affected by climate, and the cover of cloud, fog and haze is the most influence factor of fetch the information. Therefore, effective process of the image could promote the usability of image, and lower the weather conditions of aerial photography. There are a lot of methods to image dehazing. In this study, we discusses the Retinex algorithm and Dark Channel Prior algorithm which are the most new mehods in image dehazed. The Retinex is based on MSR method, and improve the color correction to correspond the human vision. The dark channel prior is based on distribution of fog to promote the calculate speed of big image. The thesis will use different images to experiment, and use the white balance as after process of image. Finally use the image quality assessment index to evaluate the dehazed-image. Keywords: Image Dehazing, Retinex Theory, Foggy Model, Dark Channel Prior