Spatial contextual feature plays an important role in high resolution remote sensing image. A new approach based on local variance analysis is introduced to spatial contextual feature extraction for change detection in this paper. In the proposed approach, the change magnitude between the paired central pixels of a local area (e.g., a 3 × 3 sliding window) in multitemporal images depends on surrounding pixels in the difference image, and this change magnitude is quantitatively measured based on the standard deviation of the difference image within the local area. This difference image contextual property may be useful particularly when high-resolution images are used. Finally, the change magnitude image is classified as a binary CD map by using SVM. This proposed approach is applied to SPOT-5 multitemporal datasets and two QuickBird multitemporal images from two case studies, which are quantitatively compared and validated. Experimental results show that the proposed approach is feasible.