|
Article Excerpt
Introduction
Human well-being is connected to features of the built environment. Although the latter is an intuitively acceptable fact, research into the nature of this relationship has turned out to be a challenging trans-disciplinary undertaking. A recent overview of studies addressing various psychosocial features of urban neighbourhoods that can affect residents is part of a study by Spokane, Lombard, Martinez, Mason, GormanSmith, Plater-Zyberk, Brown, Perrino, & Szapocznik, (2008). Behavioural survey methods employing human test subjects are naturally dependent on a variety of subject-related factors, which can affect an architectural evaluation. It was found, for example, that judgement on landscape preference can depend on the nationality (Purcell, Lamb, Peron & Falchero, 1994) or ethnicity (Kaplan & Talbot, 1988) of test subjects.
The quest for a measure which allows direct comparability and better objectivity has motivated the use of the fractal dimension for architectural image analysis (Bovill, 1996; Mandelbrot, 1983; Ostwald, Vaughan, & Tucker, 2008). A review of studies on visual perception suggests that people aesthetically prefer patterns of mid-range fractal dimensions (Taylor, 2006). This observation was corroborated by physiological pilot experiments based on skin-conductance measurements, which indicated that mid-range fractal dimensions have the most positive effect on subjects' stress (Taylor, Spehar, Wise, Clifford, Newell, Hagerhall, Purcell, & Martin, 2005). A recent investigation used quantitative electroencephalography (qEEG) to record psycho-physiological responses in the cortex of subjects viewing computer-generated fractal silhouettes which underwent controlled changes of their fractal dimension. The results of this study support the view that mid-range fractal dimensions play a unique role in visual perception (Hagerhall, Laike, Taylor, Kuller, Kuller & Martin, 2008).
Architectural image analysis often focuses on the complexity of house facades and streetscapes. However, for the aesthetic assessment of distant urban views, the complexities of city skylines are at least as important (Heath, Smith, & Lim, 2000). The skyline is the contour of the sky segment in an image and its fractal dimension is regarded as an important feature which may be used to characterise natural scenes (Hagerhall, Purcell, & Taylor, 2004; Keller, Chen, & Crownover, 1989; Keller, Crownover, & Chen, 1987). Psychological eye-tracking experiments have demonstrated that contours with high intensity gradients attract subjects' attention (Rayner & Pollatsek, 1992). Fractal analysis of urban skylines has previously been conducted by several studies (Cooper, 2000, 2003; Oku, 1990). The theory of contextual fractal fit implies that cityscapes look better if the fractal dimension of their skyline matches the fractal dimension of the environment (Bovill, 1996; Stamps, 2002).
A general aim of the present study is to understand better the impact of skyline complexity on aesthetic judgement and to develop a software tool for architectural image analysis which can help to improve comparability and objectivity. This article contributes a method which allows extraction and fractal analysis of an approximation of the skyline from digital images of cityscapes. The experimental section illustrates how to determine the skyline even in the presence of obstacles such as power-lines, poles and cranes which intersect the skyline.
In order to extract the skyline in digital images, regions which contain sky have to be identified. Specific methods for detection of sky regions have previously been developed for various applications. These include photographic image manipulation and image/video enhancement, as well as areas such as image understanding or semantic image retrieval where the sky region can help to determine the orientation of an image or to identify outdoor images. Luo and Etz, (2002) proposed a model-based sky detection method that incorporates colour classification by a multilayer perceptron and a physics-motivated sky signature validation. The sky detection approach of Gallagher, Luo and Hao, (2004) used a two-dimensional polynomial model of blue sky in an image to improve the approach of Luo and Etz, (2002). The task of real-time sky...
|
