Shadow Detection and Sun Direction in Photo Collections

Abstract

Modeling the appearance of outdoor scenes from photo collections is challenging because of appearance variation, especially due to illumination. In this paper we present a simple and robust algorithm for estimating illumination properties-shadows and sun direction-from photo collections. These properties are key to a variety of scene modeling applications, including outdoor intrinsic images, realistic 3D scene rendering, and temporally varying (4D) reconstruction. Our shadow detection method uses illumination ratios to analyze lighting independent of camera effects, and determines shadow labels for each 3D point in a reconstruction. These shadow labels can then be used to detect shadow boundaries and estimate sun direction, as well as to compute dense shadow labels in pixel space. We demonstrate our method on large Internet photo collections of scenes, and show that it outperforms prior multi-image shadow detection and sun direction estimation methods.

A geotagged image dataset with compass directions for studying the drivers of farmland abandonment

“…For example, it is important to know the position of the photographic camera when validating metadata [5] . In certain circumstances, it is possible to derive certain characteristics from the information contained in the photograph itself, such as the relative position of the sun with respect to the camera, or the approximate configuration of the camera at the time of capture [6 , 7] . To validate this type of calculation, other authors have taken advantage of data sources such as aerial photographs, Street View, or 3D models [8 , 9] .…”

Detecting and Correcting Shadows in Urban Point Clouds and Image Collections

“…In a somewhat similar manner, Wehrwein et al [15] automatically detect the shadows and the sun direction in a series of picture. Applying SfM to the pictures yields a 3D model.…”

“…The luminance correction step requires an approximation of the sunlight orientation to be able to separate the different component of the scene lighting. Although the method proposed by [15] locates the sun orientation without a priori information, it cannot be applied on large urban scans due to the lack of precision on the surface to detect attached shadows. However, the sun azimuth and elevation can be estimated if the approximate GPS position and time of the data acquisition are known [11], leading to the sun orientation estimate.…”

Simple Global Thresholding Neural Network for Shadow Detection

“…(5) On the basis of the characteristic that the intensity of shadow pixels is lower than that of non-shadow pixels, Wehrwein et al proposed an adjacent ratio map that showed the ratios of adjacent pixels of a gray image and used an empirical threshold for shadow detection. (6) Jung et al obtained a higher-order residual map in a log domain and detected shadows using an empirical threshold. (7) Newey et al found that texture features are invariant to illumination, and they adopted a local binary pattern feature and the Otsu method to detect shadows on the basis of image segmentation.…”