The traditional multi-feature fusion change detection does not consider the fact that different features contribute differently toward the change detection results. Furthermore, the traditional Markov random field (MRF) change detection quality is affected by the spatial information weight. This study proposes a novel change detection method based on density attraction and multi-scale and multi-feature fusion. First, the texture difference image is obtained by local similarity measurement and information entropy on the basis of extracting Gabor texture features, and the spectral difference image is calculated by change vector analysis. Then, the adaptive method is used to fuse the spectral and texture differences. Finally, the density attraction model is combined with the traditional MRF to construct an adaptive weighted MRF model and obtain the change map of a difference image. The experimental results show that the proposed method can not only make full use of different features, but also well maintain the image edge details and improve the change detection accuracy.