In this paper, quantized residual preference is proposed to represent the hypotheses and the points for model selection and inlier segmentation in multi-structure geometric model fitting. First, a quantized residual preference is proposed to represent the hypotheses. Through a weighted similarity measurement and linkage clustering, similar hypotheses are put into one cluster, and hypotheses with good quality are selected from the clusters as the model selection results. After this, the quantized residual preference is also used to present the data points, and through the linkage clustering, the inliers belonging to the same model can be separated from the outliers. To exclude outliers as many as possible, an iterative sampling and clustering process is performed within the clustering process until the clusters are stable. The experiments undertake indicate that the proposed method performs even better on real data than the some state-of-the-art methods.

The management and application of remotely sensed data has become much more difficult due to the dramatically growing volume of remotely sensed imagery. To address this issue, content-based image retrieval (CBIR) has been applied to remote sensing image retrieval for information mining. As a consequence of the growing volume of remotely sensed imagery, the number of different types of image-derived products (such as land use/land cover (LULC) databases) is also increasing rapidly. Nevertheless, only a few studies have addressed the exploration and information mining of these products. In this letter, for the sake of making the most use of the LULC map, we propose an approach for the retrieval of alike scenes from it. Based on the proposed approach, we design a content-based map retrieval (CBMR) system for LULC. The main contributions of our work are listed below. Firstly, the proposed system can allow the user to select a region of interest as the reference scene with variable shape and size. In contrast, in the traditional CBIR/CBMR systems, the region of interest is usually of a fixed size, which is equal to the size of the analysis window for extracting features. In addition, the user can acquire various retrieval results by specifying the corresponding parameters. Finally, by combining the signatures in the base signature library, the user can acquire the retrieval result faster.

Keywords: Multimodal Descriptor, Visible and Thermal Infrared Image Matching, Spatial Structural Information, Frequency Structural Information, Multi-spatial Regions Division Due to the differences of imaging principles, image matching between visible and thermal infrared images still exist new challenges and difficulties. Inspired by the complementary spatial and frequency information of geometric structural features, a robust descriptor is proposed for visible and thermal infrared images matching. We first divide two different spatial regions to the region around point of interest, using the histogram of oriented magnitudes, which corresponds to the 2-D structural shape information to describe the larger region and the edge oriented histogram to describe the spatial distribution for the smaller region. Then the two vectors are normalized and combined to a higher feature vector. Finally, our proposed descriptor is obtained by applying principal component analysis (PCA) to reduce the dimension of the combined high feature vector to make our descriptor more robust. Experimental results showed that our proposed method was provided with significant improvements in correct matching numbers and obvious advantages by complementing information within spatial and frequency structural information.