The Mosaic Algorithm Of Panoramic Image

The nonsubsampled contourlet transform (NSCT) has characters of multi-scale, multi-direction and shift-invariance. It is composed of nonsubsampled Laplacian pyramid and nonsubsampled directional filter banks. Firstly, multi-scale decomposit-ion of the image is performed by nonsubsampled Laplacian pyramid, then directional decomposition of the high frequency sub-bands in each scale are achieved by nonsubsampled directional filter banks, Finally, sub-band images of multi-scale, multi-direction can be got with the same size of the source image. Because of the characters of the nonsubsampled contourlet transform, it can be used in the image stitching, the result image is more natural, the details are clearer. For the B-spline surface fitting has the advantage of flexibility and the two-dimensional histogram can fully represent the information of image, the color correction of mosaic image can use the algorithm combining with the former two methods, then the color transition is more natural.This dissertation systemically study the image stitching using the non-subsumpled contourlet transform, at the same time it studies the color correction make use of two-dimensional histogram and B-spline surface fitting, The main content and achievements are as follows:1. The basic principles of fourier transform, the advantages and disadvantages in the processing of image were particularly studied. The basic principles and algorithm of the wavelet transform and 2D wavelet transform were analyzed based on the former, proposes the defects of the wavelet transform which was used in the processing of image, the basic principles,developing and application of multi-scale analysis in image at the same time.2. The basic principles and structure of contourlet transform and nonsubsampled contourlet transform were particularly studied. Compare with contourlet transform, because NSCT does not have down-sampling process, so it has the advantage of shift-invariance. For the advantage, this paper proposes a new algorithm of image stitching using NSCT. The experiment proves that the result image of this algorithm is better and the transition is natural compare with the algorithm of nonsubsampled wavelet transform.3. The problem which was caused by different light and dark scenes and digital camera auto white balance in the processing of image stitching, firstly the basic principles of color fusion algorithm and color histogram algorithm were studied, and proposes a new color correction based on 2D histogram and B-spline surface fitting. The experiment proves that the transition of color is more natural and the image is clearer compare with the former two algorithms.