A method for the visual analysis of streetscape character using digital image processing

This paper outlines an interdisciplinary approach, utilising architectural knowledge and computer imaging, to develop an analytical tool that describes the visual characteristics of a streetscape. By segmenting a digital image into regions defined by edges, the streetscape can be studied as an arrangement of regions or elements with a particular relationship Different parts of the street (usually defined a by a single dwelling) can then be related to each other to find what similarities, if any, exist at a purely visual level. The concept of the use of streetscape character as a planning instrument is typically discussed in terms of imageability and spacial configuration (Lynch 1960 and Hillier 1996). A method is developed in the present paper that outlines the digital image processing techniques used to record dwellings within a locality. The method utilises the Hough transform, an algorithm that provides a global measure of the geometry within an image. This algorithm is conceptually related to visual partitioning within Gestalt psychology as well as to the theory of description retrieval developed by Hillier. The process used to define variables within the algorithm and the way in which images of a streetscape have been collected are discussed. The method developed in this paper indicates that digital image processing can differentiate the visual composition of elements within a streetscape. The method will form the basis for future research concerning the visual characteristics of the built environment