局部亮源區域對比增強的高動態範圍影像

由於顯示器的進步，⾼動態範圍的影像（ＨＤＲ）在最近變得越來越熱門，但是在高動態範圍影像的擷取上，專門用來擷取高動態範圍的相機對於⼤部份的⼈來說都過於昂貴，因此有個反色調映射算法（ＩＴＭＯ），它是將低動態範圍的影像利非線性或線性的曲線映射到⾼動態範圍去產生高動態範圍的影像，雖然大部份的反色調映射算法可以產生不錯的⾼動態範圍影像，但是仍然會產生一些問題，我們將問題分類成細節的損失問題和不自然的光暈問題，在這篇論文，我們提出了局部亮源區域對比增強圖去解決亮度過飽和的問題並保留細節資訊，我們結合兩張高斯模糊圖去產生我們提出的對比增強圖，一張是為了做邊緣模糊，另一張是為了產生亮源區域的細節，我們使用偽多重曝光反色調映射演算法去產生多張不同曝光的高動態範圍影像，分別是最暗, 暗, 普通, 亮, 最亮，再利用不同曝光的影像去產生不同區域清晰的細節，我們可以得到一張高品質的高動態影像，接著將最暗曝光的高動態影像和高品質的高動態影像與對比增強圖做結合，可以避免過飽和的問題並且清楚的保留了亮源區域的細節。跟其他的方法比較，我們的方法在相似性上比 Pseudo-Multiple-Exposure HDR 方法好上1.7 ％，在視覺化差異量測75 ％和95 ％分別比Pseudo-Multiple-Exposure HDR 方法好上67 ％和69 ％，在平均對比改變錯誤上⽐Pseudo-Multiple-Expsure HDR 方法好上62 ％。

關鍵字

高動態範圍；反色調對映；對比增強圖；細節損失問題；不自然的光暈問題

並列摘要

The High dynamic range (HDR) imaging becomes more and more popular recently due to the advance of display devices. However, the acquisition of HDR images from HDR cameras is expensive for most people. The HDR inverse tone mapping (ITMO) is a technique, that maps low dynamic range (LDR) images to HDR images. Although most ITMO method can produce good HDR images, however, there are two problems in specular or light source regions of the HDR images: loss of detail and unnatural Halo effects. In this thesis, we propose a contrast enhanced expand map to reduce the over saturation problems for preserving the details information. We fuse two Gaussian maps to generate an expand map. One Gaussian map is for edge blurring, the other Gaussian map is to recover details of the specular area. We adopt a pseudo multiple exposure ITMO to generate multiple HDR images with different exposures. The darkest emulated HDR image and the expand map are used to prevent saturation and reveal the details in the specular areas. According the simulation results, our proposed method can reduce the over saturation problem and reveal details in specular areas. It also reduces the halo effects in light source areas. Comparing with Pseudo-Multiple-Exposure HDR method, the proposed method has better average values in SSIM by 1.7 ％ . The proposed method has better average values in HDR-VDP-2 75 ％ and 95 ％ by 67 ％ and 69 ％. The proposed method has better average values in mean conrast error by 62 ％.

並列關鍵字

HDR ； HDR ITMO ； Contrast enhance map ； over saturation ； unnatural halo effects

參考文獻

imaging and low dynamic range expansion for generating HDR content.”

Computer graphics forum. Vol. 28. No. 8. Blackwell Publishing Ltd, 2009.

tone mapping of legacy video and photographs.” ACM Transactions on

[3] G. Guarnieri, S. Marsi and G. Ramponi, “High Dynamic Range Image Display

With Halo and Clipping Prevention,” IEEE Transactions on Image

國際替代計量

局部亮源區域對比增強的高動態範圍影像

全文下載

主題瀏覽