Earticle

This paper presents a novel remote sensing image fusion algorithm, which implements the intensity-hue-saturation (IHS) transform on panchromatic sharpening of multispectral data and the dual-tree compactly supported shearlet transform (DT CSST) during fusion. Shearlet transforms can provide almost optimal representation of the anisotropic features of an image. The spatial domain discrete implementation, the compactly supported shearlet transform (CSST), which represents the directions by dilation operations, are selected in the proposed fusion method. Since most of the prominent features of images, such as edges and regions, have limited sizes in the spatial domain, CSST is very suitable for image fusion. However, the conventional CSST is shift-variant, which causes distortions in fused images. With the embedded dual-tree (DT) sturcture in the CSST, the shift-variant properties can be effectively reduced. Combining the IHS transform and the DT CSST, an effective panchromatic and multispectral image fusion method is proposed in this paper. The experiments’ results suggest that the proposed method extract more spatial information from panchromatic images with less lost in spectral consistency compared to other fusion methods which are based on discrete wavelet transform (DWT), à trous wavelet transform, à trous shearlet transform, the dual-tree complex wave transform ( DT CWT), or the Curvelet transform.

In order to decrease the smoothing effect of traditional linear interpolation methods and avoid the complexity of edge-directed adaptive algorithms, a novel image interpolation algorithm based on the directional extension of linear interpolation kernel is proposed. At first, All Phase Non-subsampled Contourlet Transform is used to divide the image of low resolution into different frequencies and directional sub-band images, and then the All Phase Discrete Cosine Transform Interpolation Kernel is extended to 2-D with distinct directional priorities and used to enlarge respective sub-band images. The interpolated sub-band images are finally synthesized as the upsampled image. Simulation results demonstrate that the proposed algorithm improves both the subjective and objective quality of the interpolated images over conventional linear interpolation. The computation complexity of the proposed algorithm is moderate and it can be implemented with parallel structure so as to save more computation time.

The detectors are an important tool in deep space exploration; they can provide a large number of images which contain the ground environment information of celestial body. Bythree-dimensional reconstruction, terrain information can be extracted from these images, such as elevation data and texture data. This paper focuses on an adaptive terrain reconstruction for sequence image. A method, which calculate Binocular geometric model with matching measures of matching points and random sample consensus algorithm, is proposed. On the basic of this method, a densepoints matching method based on epipolar and homography constraints is presented. Besides the selection method of reconstruction image and the terrain reconstruction flow are discussed. The proposed adaptive terrain reconstruction satisfies the automation and real time requirement of some terrain reconstruction task, such as landing filed selection in detector landing, and is experiment and applied into sequence image of detector ChangE III’s landing camera, so as to ensure the safety of detector landing.

Through the analysis on the unique characteristics of Uyghur characters, in order to further improve the recognition rate, this paper developed the Center Distance Feature (CDF) to its modified form which is named as Modified Center Distance Feature (MCDF). By combination with some low dimensional features including stroke number feature, additional part’s location feature, shape feature, bottom-up and left-right density feature(BULR) in experiments, MCDF gifted robust recognition accuracy of 98.77% for the 32 isolated forms of Uyghur characters. MCDF increased the recognition accuracy by 4.51 points comparing with the result from the combination of CDF with the same low dimensional features mentioned above, which is 94.16%. This paper used the samples from 400 different volunteers. The recognition system is trained using 70 percent of 12800 samples from 400 different writers and tested on the remained 30 percent.