The efficiency of change detection methods, in terms of content discrimination, is degraded by the presence of noise. Furthermore, illumination changes tend to cause further degradation in change detection accuracy by swamping content changes of similar magnitude. This fact imposes the application of a method which detects illumination variations in the presence of occlusions. This paper proposes the application of a block clustering method which aims to separate content changes from noise-level changes. The algorithm is performed in conjunction with a brightness normalization technique for the correction of ambient illumination variations.