In the processing of registration for multi-temporal high resolution remote sensing images, the phenomena of surface features change and relative parallax displacement caused by differences in acquisition conditions degrades the accuracy of registration. To resolve the aforementioned issue, a registration algorithm for multi-temporal high resolution remote sensing images based on abnormal region sensing was proposed, which consists of coarse and fine registration. The algorithm of Scale-Invariant Feature Transform (SIFT) has a better performance on scale space, the feature points from different scale space indicates the various size of spot. The high scale space points represent the objects which have a stable condition, the coarse registration can be executed depending on those points. For the fine registration, intensity correlation measurement and spatial constraint were used to decide the regions which were used to extract the efficacious points from low scale space, the areas for searching matching points were limited as well. Finally, the accuracy of the proposed method was evaluated from subjective and objective aspects. Experimental results demonstrate that the proposed method can effectively restrain the influence of abnormal region and improve registration accuracy.

Since the contradiction between edge-preserving in homogeneous regions and structure-preserving in the boundary region of an image, a new algorithm combined with cross-based framework and guided filter was proposed. The main idea of the algorithm was adding an adjust offset in guided filter to ensure remaining the edge structure. Usually the fixed size window was used as neighborhood filtering, while the new algorithm employed cross-based framework, which chose a threshold in grayscale similarity. Taking the advantage of stereo matching, the adaptive filtering blocks whose sizes and shapes can adjust automatically were generalized. The adjust offset was proportional to the threshold, which was more robust than a hard threshold. In the simulation experiments of processing international standard sequence, the blocks were generalized by cross-based framework efficiently and effectively, homogeneous regions were smoothed well. The added offset outperforms many other algorithms in terms of sharpness enhancement. Compared to the guided filter, the value of Peak Signal-to-Noise Ratio (PSNR) of the proposed method has been improved by about 2 dB. The test results of real natural picture show that the proposed algorithm has a good future in practical application.

Concerning low classification accuracy and high false alarm rate of current intrusion detection models, an intrusion detection classification model based on Landmark ISOmetric MAPping (LISOMAP) and Deep First Search (DFS) Relevant Vector Machine (RVM) was proposed. The LISOMAP was adopted to reduce the dimension of the training data, and RVM based on the DFS was used for classification detection. Compared with the Principal Components Analysis (PCA)-Supported Vector Machine (SVM), the experimental results indicate that the LISOMAP-DFSRVM model has lower false alarm rate with almost the same detection rate.