Obtaining accurate change detection results from high-resolution satellite sensors

Change detection for high resolution satellite images, based on SIFT descriptors and an a contrario approach

“…The GeoCDX method [2] extracts features such as entropy, skewness and lines, and operates a pixel-based change detection with a spatial tolerance to misregistration. Bryant et al [3] perform a background characterization from the multi-spectral bands of the image, followed by a spectral differentiation and a detection of change pixels with a Bayesian classifier. Gamba et al [4] use a feature-based change detection technique (with edge detection and matching) to correct misdetections, due to misregistrations or miscalibrations, from an area-based change detection map.…”