그림자 영역의 수직 경계선을 이용한 고해상도 위성 영상의 그림자 제거

Abstract

The effect of shadows in very-high resolution (VHR) satellite imagery has become one of major issues in urban area analysis. The shadows in VHR satellite imagery, such as Kompsat-3A, covers a wide range of urban areas due to the buildings and affects the analysis of land cover and land use. This paper presents a shadow-removal method using the linear relationship between inner and outer intensities of shadow region borders. Vertical shadow borders, which have constant orientation in the whole image according to the altitude and azimuth of the sun at the image acquisition time, are extracted from shadow regions based on the orientation and linearity of the vertical border pixels. We estimate the pixel intensities in the shadow region based on RANdom Sample Consensus (RANSAC)-based linear transformation using only the pixels belonging to the shadow and non-shadow regions located outside a narrow border region. The pixel intensities in the narrow border region are interpolated using the least-square method which calculates the intensity plane parameters using the pixels belonging to small shadow and non-shadow regions. The proposed shadow removal method was tested using a part of WorldView-3 satellite images and showed improved shadow removal performance than the conventional linear correlation correction method in terms of image contrast and detailed-information improvement in the restored shadow region.