A Study on Shadow Detection and Color Restoration in Aerial Images

空照影像陰影區塊之偵測和色彩修復之研究

張志瑋(Chih-Wei Chang) ； 蔡展榮(Jaan-Rong Tsay)

《航測及遙測學刊》 21卷4期 (2016/12) Pp. 239-256

https://doi.org/10.6574/JPRS.2016.21(4).3

摘要

The intensity value on each image pixel is determined by the functions of incident illumination and reflection of ground objects. In fact, light cast on terrain objects such as buildings and trees would cause shadows and occlusion. It is hard to restore original textures and color information in shadow areas. Many detailed information is hidden or lost due to such problems. Moreover, false color tone, shape distortion and inaccurate image matching within shadow areas also infect image recognition. This paper proposes an algorithm for automatic shadow detection based on the principal component analysis(PCA). It is tested by using both RGB and HSI bands in order to determine a proper and suitable color space for shadow detection. Besides shadow detection, color restoration in shadow areas is an important issue, too. This paper also performs subjective and objective evaluation on both histogram matching method and the local statistics method for color restoration in shadow areas. Test results demonstrate satisfied quality for restored images in shadow area.

關鍵字

shadow detection ； color restoration ； principal component analysis ； histogram matching ； local statistics method ； aerial images

並列摘要

影像中每個像元的灰階值是由光照及地面反射函數來決定，當陽光照射地表時，較高的地表物(例如建物、高樹等)或大氣中的雲層將光線遮蔽使得亮度發生明顯變化而造成陰影的產生，陰影區內地物的影像資訊量因此降低甚至遺失，陰影亦會造成影像匹配失敗、不自然的顏色變化、錯誤的形狀變形等問題以及影響影像辨識。本研究提出了一個以影像主成分分析為基礎的自動化陰影偵測方法，此方法對RGB 和HSI 兩種色彩空間的結果進行比較，嘗試找出最合適與有效的色彩空間。除了陰影的偵測，陰影區的色彩修復也是很重要的一環。雖然陰影區內的資訊被降低，但陰影區附近的影像依然提供一定程度的亮度及色彩資訊可供陰影區色彩修復之用，本文除了偵測陰影外，亦對常見的兩種陰影修復方法進行主觀和客觀的評估：直方圖匹配以及局部統計法，結果顯示陰影區內的量化細節以及令人滿意的修復成果。

並列關鍵字

陰影偵測 ； 色彩修復 ； 主成分分析 ； 直方圖匹配 ； 局部統計法 ； 空照影像