Research On Remote Sensing Image Fusion Algorithm Based On Nonsubsampled Contourlet Transform

Scholars have been concerned about wavelet transforms in the direction of one-dimensional data, because it can reflect a zero dimensional singular points. Though in the structure of an image with lines and surfaces, the wavelet transform doesn’t seem to be a good acquisition to the edge of the image texture, multiscale analysis appears well to fill the gaps. Since being anisotropy, Contourlet transform not only have the frequency characteristics of wavelet transform, but also contains special directional characteristic. Compared to the wavelet transform, the latter can use less coefficient to express image information after conversion. Contourlet perform well in terms of gathering edge information. Evolved in the theory of CT transform, NSCT transform is an effective method. It overcomes the translation invariant features of CT transformation.It is the image of redundant decomposition, ie, map and transform the resulting sub-band source image data of equal size, so that the edge of the NSCT transform can retain more images and texture information, and ultimately to be more accurate restoration of. the original source image.This paper studies the application of polarization images of CT transformation and SAR imaging applications in NSCT Transformation. The first part focuses on the polarization image after the CT of the high-frequency coefficients of the decomposition process. After CT transformed, polarization image will be into a series of low frequency sub-frequency coefficients and coefficients of different sub-graphs for different directions under the scales. In this paper, we use the law of averages to deal with the low-frequency coefficients. For the high-frequency coefficients, this paper uses the absolute value of the pixel values data, regional energy law, region energy maximum, maximum standard deviation, maximum corrcoef, improved correlation coefficient method, whichever is greater regional energy and improve the correlation coefficient binding assay and regional energy whichever is greater value to the maximum horizontal gradient method. The fourth chapter focuses on the study of SAR image fusion in NSCT transform. For the decomposition of the low-frequency sub-band coefficients, we use the law of averages to select the appropriate pixels; For high frequency subband coefficients, we select the fusion rule to choose the gradient greater regional value consistent with the regional comparison. Comparing with LP transform and WT transform, the algorithm is the best on running the algorithm in Matlab. And the SPOT image and TM image fusion in remote sensing research, this section mainly discusses the use of non fusion and decomposition Nonsubsampled Contourlet transform multi-scale multi direction SPOT image and TM image fusion algorithm, the fusion region statistical features.