延伸曝光曲線線性特性之調適性高動態範圍影像融合演算法

摘要

因為影像感測器的動態範圍限制，拍攝的數位影像時常與實際的視覺感受有所差異。為了讓產生的影像可以顯示實際環境的亮度分佈，產生高動態範圍(HDR) 影像的技術應該是較佳的解決方案。高動態範圍影像提供了較以往傳統數位影像更廣的動態範圍，且可以精準的重現真實景象裡的光強度分佈。

本研究提出一套調適性高動態範圍的影像融合演算法。運用影像感測器曝光特性曲線，及每張影像的曝光參數，可以將實際光強度正確的對應到數位影像資訊。透過有效的融合不同曝光條件的影像的方法，可以延伸圖像光強度的線性區間。基於保留影像資訊的原則，對影像亮度資訊作處理，以便於在影像中準確地重現真實景像中的光強度分佈，此外，影像的邊緣強度資訊也將在此處理程序被保留。

實驗結果顯示，本研究提出之演算法較現行的方式，可以達到較廣的動態範圍。不論在主觀或客觀測試上，都有更較佳的效果。在不同的拍攝環境下，本演算法產生的高動態範圍影像在較亮以及較暗區域的細節與對比皆有較好的表現。

　

Adaptive high dynamic range image fusion algorithm based on Extended Linear Exposure Characteristics

Abstract

The photo captured by the camera is often different from what we see due to the limited dynamic range of the camera sensors. To produce an image with true luminance, the high dynamic range (HDR) image capturing technique is a good solution. It provides a wider dynamic range than the traditional techniques, and is able to re-produce the luminance with high fidelity.

This work proposes an adaptive high dynamic range image fusion algorithm. By utilizing the exposure curve characteristics of image sensors and the exposure parameters of each image, it can convert the actual luminous intensity into image data with high accuracy. This method can extend the linear range of the luminous intensity by effectively fusing images with different exposures. The luminance processing is based on the principle of preserving the image information so as to re-produce the true luminous distribution of the image. In addition, the edge information is also kept with the propose method.

The simulation results show that the proposed algorithm is able to reach wide system dynamic range. Both objective and subjective measures show superior performance compared with existing methods. With different capturing environments, it generally produces high dynamic range images with better performance of details and contrast in both bright and dark areas.