Estimation of ambient illumination variation between colour images in the presence of content changes for real-time illumination-invariant change detection

摘要

The task of detecting changes between two image frames is obstructed by the influence of noise and by the existence of ambient illumination variations between the image frames. The former is an inherent property of all electronic imaging devices. The latter appears when changes in camera exposure or white balance settings occur and tends to degrade the efficiency of change detection, if left untreated prior to frame differencing. An additional difficulty lies in the fact that a change detection process is required to accurately detect illumination changes in the presence of both luminance and content changes. In this paper, a luminance invariant change detection method is proposed. It consists of a three-channel brightness correction stage employed for the brightness normalization process and followed by a block-based clustering method, which aims to detect content changes from changes caused by the influence of noise.