Image Edge Detection And Coding Based On Contourlet

The limitations of commonly used 2-D transforms,as separable extensions of one-dimensional transforms,such as the Fourier and wavelet transforms,in capturing the geometry of image edges are well known.The aim of Multiscale Geometric Analysis(MGA)is to find a kind of optimal representation of high dimension functions in the sense of nonlinear approximation,which can capture the intrinsic geometrical structure that is the key in visual information.Contourlet transform is a new MGA method which has lower redundancy and computation complexity.Especially it can represent the 2-D singularity signals effectively.It has become a hot topic of researches by the scholars worldwide.This paper applies Contourlet transform to image edge detection and compression.New schemes are proposed based on modified Contourlet transform and other improved methods.The major contribution of this dissertation is stated as follows:(1)To overcome the Contourlet’s lack of translation invariance which could cause visual artifacts by setting some transform coefficients to zero,a method for image edge detection based on Non-Subsampled Contourlet Transform(NSCT)is researched.(2)An adaptive multiscale edge detection method based on Contourlet transform is proposed.In this method,the multiresolution technology is used so that the image can be analyzed from coarse to fine.Therefore the newly proposed edge detection method outperforms the wavelet operator in both resistant of noise and false edge.(3)Contourlet transform has good nonlinear approach ability,but its redundancy is a block for image comppression.To overcome the redundancy of Contourlet transform,the laplacian pyramid(LP)filter banks are discarded which could induce redundancy.By using the wavelet for multiresolution decoposition and then add Directional Filter Banks(DFB)to high frequency subbands for directional decoposition,the Wavelet Based Contourlet Transform(WBCT)is researched.(4)Since WBCT can bring visual artifacts to the smooth area of an inage,it is optimized by using Multiscale Directional Filter Banks(MDFB)which has better capability instead of DFB,hence the transform of Wavelet-Based Multiscale Directional Filter Banks(WMDFB)is researched.It can suppress the visual artifacts occered in WBCT effectively.(5)At low bit rate,most computation for Post-Compression Rate-Distortion Optimization(PCRD-opt)in Tier 1 of JPEG2000 is redundant.A novel Successive-Approximation PCRD-opt(SARD-opt)rate control method is proposed to alleviate this problem.(6)PCRD-opt algorithm yields a rate-distortion optimal representation of the image,achieves quality layers and offers excellent distortion scalability.Benefited from the modified significance context model,the MQ encoder in the WMDFB encoder provides the output which bit-rates approximate their entropy-rates more efficiently,while JPEG2000 context model offers poor initial probability estimation for the input of the MQ encoder.Results show the proposed approach outperforms the WMDFB block coding method based on JPEG2000 context model and the JPEG2000 method,especially for the images that possess abundant textures.