| Beyond wavelet transform is a new multiscale analysis based on the wavelet ananlysis. It overcomes the disadvantage of traditional two-dimensional wavelet transform in the direction, sparseness, etc. At present, the research and application of beyond wavelet is development, some of the commonly used are ridgelet, curvelet, surfacelet, wedgelet, beamlet and contourlet. Smoothlet transform is proposed as horizon approximation of wavelet transform for curved line based on second-order-wedgelet. It can be a good approximation for horizon. But the approximation approach is rough and the conversion efficiency is low. In this paper, we researched the image coding theory of N-smoothlets based on those of smoothlet. The main research content is as follows:1. Improved the Smoothlet transform. In view of the approximation is low and the transform efficiency is low for traditional smoothlet transform. In this paper, improving approximated method is presented based on smoothlet transform. The improving method transit the mutation pixels existed in overlapping parts and filter the mutational pixels result from the determine order of approximated curve. The algorithm is in the guarantee of the reconstruction image quality and improves the efficiency of transformation.2. Improved the image coding theory of Smoothlets. Make corresponding improvement for the coding theory of N-smoothlets based on the coding theory of smoothlet. The simulation result indicated that N-smoothlets is better to realize the image compression coding compared with smoothlet when N=3, and better than the JPEG2000, in low bites.3. Put forward the algorithm of removing the block effect for N-smoothlets reconstructed image. In generally, N-smoothlets transform segment the image into several independent macro block to approximate. The reconstruction image has block effect because of ignoring the correlation between block and block. Put forward an adaptive algorithm to eliminate the block effect according to the strength of the block effect between different image block. The algorithm reasonable adjust the number of pixels to filter and the choice of the number of the filtering pixels is with the method of coarse and fine tuning to determine the value of r. The algorithm can have goodremoval efficiency for the regional block effect and can be a very good protection of image texture information, and then improve the efficiency of transformation. |
