Detecting regions of change in multiple images of the same scene taken at different times is of widespread interest due to a large number of applications in diverse disciplines, including remote sensing, surveillance, medical diagnosis and treatment, civil infrastructure, and underwater sensing. This paper presents a systematic survey of the common processing steps and core decision rules in modern change detection algorithms, including significance and hypothesis testing, predictive models, the shading model, and background modeling. We also discuss important preprocessing methods, approaches to enforcing the consistency of the change mask, and principles for evaluating and comparing the performance of change detection algorithms. It is hoped that our classification of algorithms into a relatively small number of categories will provide useful guidance to the algorithm designer.

Introduction The recent developments in satellites and remote sensors, together with the necessity for an efficient control of the environment (management of natural resources, risk assessment, damage mapping, land use monitoring, ...), offer new challenging applications. Multitemporal change detection is one of them and a number of methods has been designed the last few years (e.g. [1--3]). Basically, change detection techniques rely on some clustering schemes that identify the coordinates of pixels that have changed between two dates. In this work, we are concerned with SAR images and we assume that the images have been geometrically corrected and co-registered. An exemple is given with the three SAR images in figure 1. They show a rice plantation in Semarang (Java Island) with mainly early rice, late rice and other (a) ##### - Feb. 6, 1994 (b) ##### - Feb. 16, 1994 (c) # ### - Mar. 6, 1994 Figure 1. Three co-registered ERS-1 images of a rice plantation in Java Island, Indonesia

Identifying interesting changes from a sequence of overhead imagery—as opposed to clutter, lighting/seasonal changes, etc.—has been a problem for some time. Recent advances in data mining have greatly increased the size of datasets that can be attacked with pattern discovery methods. This paper presents a technique for using predictive modeling to identify unusual changes in images. Neural networks are trained to predict “before” and “after” pixel values for a sequence of images. These networks are then used to predict expected values for the same images used in training. Substantial differences between the expected and actual values represent an unusual change. Results are presented on both multispectral and panchromatic imagery.

Abstract — This paper evaluates the potential interest of using bivariate gamma distributions for image registration and change detection. The first part of the paper studies estimators for the parameters of bivariate gamma distributions based on the maximum likelihood principle and the method of moments. The performance of both methods are compared in terms of estimated mean square errors and theoretical asymptotic variances. The mutual information is a classical similarity measure which can be used for image registration or change detection. The second part of the paper studies some properties of the mutual information for bivariate Gamma distributions. Image registration and change detection techniques based on bivariate gamma distributions are finally investigated. Simulation results conducted on synthetic and real data are very encouraging. Bivariate gamma distributions are good candidates allowing us to develop new image registration algorithms and new change detectors. Index Terms — Multivariate gamma distributions, correlation coefficient, mutual information, maximum likelihood, image registration, image change detection. I.

Vulnerable plaques are dangerous atherosclerotic lesions that bear a high risk of complications that can lead to heart attacks and strokes. These plaques are known to be chronically inflamed. The vasa vasorum (VV) are microvessels that nourish vessel walls. Proliferation of VV is part of the "response to injury" phenomenon in the process of plaque formation. Recent evidence has shown strong correlations between neovessel formation and macrophage infiltration in atherosclerotic plaque, suggesting VV density as a surrogate marker of plaque inflammation and vulnerability. We have developed a novel method for imaging and analyzing the density and perfusion of VV in human coronary atherosclerotic plaques using intravascular ultrasound (IVUS). Images are taken during the injection of a microbubble contrast agent and the spatiotemporal changes of the IVUS signal are monitored using enhancement-detection techniques. We present analyses of in vivo human coronary cases that, for the first time, demonstrate the feasibility of IVUS imaging of VV.

The statistical techniques of multivariate alteration detection, maximum autocorrelation factor transformation, expectation maximization, fuzzy maximum likelihood estimation and probabilistic label relaxation are combined in a unified scheme to classify changes in multispectral satellite data. An example involving bitemporal LANDSAT TM imagery is given.

To develop better image change detection algorithms, new models able to capture all the spatio-temporal regularities and geometries seen in an image pair are needed. In contrast to the usual pixel-wise methods, we propose a patchbased formulation for modeling semi-local interactions and detecting occlusions and other local or regional changes in an image pair. To this end, the image redundancy property is exploited to detect unusual spatio-temporal patterns in the scene. We first define adaptive detectors of changes between two given image patches and combine locally in space and scale such detectors. The resulting score at a given location is exploited within a discriminant Markov random field (DRF) whose global optimization flags out changes with no optical flow computation. Experimental results on several applications demonstrate that the method performs well at detecting occlusions and meaningful regional changes and is especially robust in the case of low signal-to-noise ratios. 1.

Automated image analysis algorithms are presented for detection and classification of changes in longitudinal time-series of color retinal fundus images. They are applicable to clinical practice, quantitative scoring of clinical trials, computer-assisted reading centers, and training. This work focuses on diabetes-related changes, although the techniques have broader applicability. Retinal features, including the vasculature, vessel branching/crossover locations, optic disk, and fovea are extracted automatically. The images are registered to sub-pixel accuracy using a 12-dimensional mapping that accounts for the unknown retinal curvature and camera parameters. The images are corrected for non-uniform illumination using a robust homomorphic surface fitting algorithm. The changes in non-vascular regions are segmented using an algorithm that is robust to relevant artifacts such as dust particles in the optical path. They are classified into five clinically significant categories using a Bayesian algorithm constrained by Markov Random Fields. A flicker animation overlaid with change analysis results allows qualitative and quantitative assessment by the user. A multi-observer validation on 43 image pairs from 22 eyes involving non-proliferative and proliferative diabetic retinopathies, showed a 96.83 % change detection rate, a 3.17 % miss rate, and a 17.65 % false alarm rate. The performance in correctly classifying the changes was 97.39 %.

Abstract — Utilizing boundaries of segmented objects from a later temporal image to constrain the segmentation of an earlier coregistered image enables information about the spectral, textural, and other characteristic attributes of image segmented objects within the two images to be mined for differences that would be indicative of specific types of land use and land cover change. Significant changes in homogeneity, hue, and vegetation indices among others provide strong cues about changes that may have occurred within segmented objects. Depending on the nature of the initial segmentation and the degree to which it was designed to extract class features of a desired size, shape, color, and texture, the method described enables highly targeted change detection to be conducted to explore desired types of land use and land cover change. For a collection of precision orthorectified QuickBird bi-temporal images, segmentation results for later images are utilized to constrain the segmentation of earlier images. Object attributes of the segmented images that provide a feature space for defining class memberships functions are employed to determine areas that were changed Keywords-change detection, data fusion, object-based classification I.

Abstract — This paper shows how to obtain a binary change map from similarity measures of the local statistics of images before and after a disaster. The decision process is achieved by the use of a ν-SVM in which a stochastic kernel has been defined. Stochastic kernel includes two similarity measures, based on the local statistics, to detect changes from the images: 1) A distance between maginal probability density functions (pdfs) and 2) the mutual information between the two observations. Distance between marginal pdfs is evaluated by using a series expansion of the Kullbak-Leibler distance. It is achieved by estimating cumulants up to order 4 from a sliding window of fixed size. Mutual information is estimated through a parametric model that is issued from the copulas theory. It is based on rank statistics and yields an analytic expression, that depends on the parameter of the copula only, to be evaluated to obtain the mutual information. Preliminary results are shown on a pair of Radarsat images acquire before and after a lava flow. A ground truth allows to show the accuracy of the stochastic kernels and the SVM decision. I.