- 學位論文
偽多重曝光並增強高亮度區域之高動態範圍影片處理
High Dynamic Range Video Processing Using Pseudo-Multiple-Exposures with Local Specular Region Enhancement
若您是本文的作者，可授權文章由華藝線上圖書館中協助推廣。
摘要
當高動態範圍顯示器成為未來的趨勢，高動態範圍影片的需求將增加，但現 有的影片大多為傳統的低動態範圍影片。為了迎合此趨勢，我們提出一個偽多重 曝光的區域增強演算法來利用低動態範圍影片產生高動態範圍影片，利用產生偽 多重曝光的影像並結合不同區域清晰的細節，我們可以得到一高品質的高動態範 圍影片。我們也提出了一個亮光圖並與最暗偽曝光的影像作內插來進一步增強高 亮度的區域。 跟不同方法的比較並作定性分析的結果顯示我們的方法在高動態範圍亮度 有較平穩的表現並能避免影片閃爍的問題，並且我們提出的方法可以在很短的時 間內產生視覺上高品質的高動態範圍影片，根據與參考圖比較得出的對比變化值， 我們的值為最小的。我們產生的亮光圖的速度為Rempel 等人的方法的78 倍快， 並且是Banterle 等人的方法的286 倍速快。
並列摘要
As the high dynamic range displays becoming the future trend, the need for HDR videos increases. However, most videos nowadays are the legacy low dynamic range (LDR) videos. In order to meet this trend, we proposed a region-based enhancement of the pseudo-exposures to generate a HDR video out of an existing LDR video. By the generation of pseudo-multiple-exposure images and the fusion of different regions with clear details, we can get a high-quality HDR video sequence. We also proposed a specular map to further enhance the high-luminance areas by interpolate with the darkest pseudo-exposure image. We compare our method with other inverse tone mapping operators, and several evaluations show that our method has a more stable performance in HDR luminance and avoids flickering artifacts. Also, our proposed method is able to generate visually high quality HDR videos with a small amount of time, which could be shown in the smallest value of overall contrast change comparing to the reference images. The generation of our specular map is 78x faster than Rempel et al.’s expand map and 286x faster than Banterle et al.’s map.
參考文獻
6501 in Proceedings of SPIE - International Society for Optical Engineering,
and Kadi Bouatouch, Image display algorithms for high and low dynamic
Wolfgang Heidrich, High Dynamic Range Imaging - Acquisition, Display,
[4] H. Seetzen, L. Whitehead, and G.Ward, A high dynamic range display using
Observations of luminance, contrast and amplitude resolution of displays,"