基於物理特性的光源估測與亮度導向色彩模型之影像除霧演算法

Abstract

在有霧的天氣中，影像品質通常會降低。在本論文中，我們提出了一種基於變分的影像去霧流程，該流程包括具有物理特性的光源估測、亮度導向色彩模型和透射率優化程序，可以有效解決這個除霧問題。我們首先提出了一個新的日/夜分類器來區分輸入影像的場景類型。然後，我們設計了一個新的照明模型，以更好地解決場景中的非全局大氣光問題。此外，我們引入了一種基於視網膜皮層理論的優化流程來獲得環境照明，且保持了場景的結構。我們在色彩模型中使用輸入圖像作為照明的初始猜測，使得夜間的顏色一致性得到保證。而在白天場景中，我們使用輸入圖像的最大通道作為照明的初始估計，使得白天黑暗區域的外觀較為自然。我們亦開發了一種基於變分的流程來平滑估計的傳輸率圖，透過該程序可以消除塊狀效應和光暈。所提出的基於亮度的校正機制在存在大片天空區域的情況下進一步提高了影像的視覺品質。我們使用現實世界的有霧影像進行實驗。綜合實驗表明，與其他最先進的算法相比，所提出的方法可以有效地提供顏色一致性、保留細節並減少結果圖像中的光暈偽影和噪聲。

Image quality is often reduced in hazy weather. In this thesis, we propose a robust variation-based image dehazing flow with a physically valid illumination estimator, a luminance-guided coloring model, and a transmission refinement procedure to effectively address this problem. We first propose a new day/night classifier to distinguish the scene type of the input image. Then, we design a new illumination model to better address the non-global air-light issue in the scene. Furthermore, we introduce a structure-preserving optimization flow based on Retinex theory to obtain ambient illumination. Color consistency in the nighttime is guaranteed because we use the input image as the initial guess of illumination in our coloring model. The natural appearance in the dark region for the daytime scene is promised because we use the maximum channel as the initial estimation of air light. A variationbased procedure is developed to smoothen the estimated transmission map, where the block effect and the halos can be eliminated through the procedure. The proposed luminance-based correction mechanism further improves visual image quality in the presence of a large sky region. Our experiments are implemented based on actual hazy images. The comprehensive experiments indicate that the proposed method can effectively provide color consistency, preserve details, and reduce halo artifacts and noise in the resulting images compared to other state-of-the-art algorithms.