Font Size: a A A

The Image Compression Method Based On Non-separable Wavelet Transform

Posted on:2017-01-14Degree:MasterType:Thesis
Country:ChinaCandidate:Y X ZhangFull Text:PDF
GTID:2348330485981324Subject:Computer application technology
Abstract/Summary:PDF Full Text Request
With the rapid development of internet and multimedia technology,massive amount of information is growing.The image which is closely related to our work and life is the most important carrier in multimedia technology.Because representing an image requires a huge amount of data,there is a contradiction between massive data and limited storage space and network bandwidth.Although the network bandwidth and storage capacity can be increased to alleviate the contradiction,the rapid growth of image data makes the investment in these two areas can' t solve the problem fundamentally.Therefore,effective compression of image data is the key to alleviate this contradiction.Image compression is achieved by removing the redundancy in the image data.There are many compression methods,and among them,the compression method based on wavelet transform has been developed rapidly in recent years.The wavelet transform appears after Fourier transform.Because of better time-frequency analysis and multi-resolution analysis which is in line with the human visual system,wavelet transform develops very fast and overcomes DCT shortcomings of square-piece effect and improves the quality of the reconstructed image,it is widely used in the field of image compression.Among various coding techniques based on wavelet transform,the coding method based on zero tree structure is most commonly used.The first classic encoding algorithm based on zero tree structure is proposed by Shapiro et al.,and it is named EZW encoding algorithm.EZW effectively uses the similarity between the wavelet coefficients,it is a simple and efficient encoding algorithm,many of the subsequent coding algorithms are based on EZW.Many kinds of compression algorithms are based on 2-D separable wavelet which equates to tensor product wavelet.Because the limited type and weak direction are defect of separable wavelet,the research on non-separable wavelet has gradually become the focus of attention.Compared with separable wavelets,non-separable wavelet has its advantages.Firstly,image is a 2-D signal,it analyses the image as a whole and can get more information.Secondly,2-D non-separable wavelet with symmetry can be used to reconstruct the image without distortion.At last,construction for non-separable wavelet is not restricted by 1-D wavelet.Firstly,this paper presents a method of constructing 2-D non-separable wavelet filter banks with orthogonal,compact support and symmetry.The method is suitable for the EZW algorithm.The classical EZW algorithm is based on separable wavelets,but only Haar wavelet has symmetry in Daubechies series wavelet.And the non-symmetrical wavelet distortion can be produced when the image is reconstructed.Therefore,the structure of the 2-D non separable wavelet filter banks is very important in image compression.Secondly,this paper presents an image compression method that is the EZW algorithm based on separable wavelet is extended to the case of non-separable wavelet transform.The method is based on 2-D non-separable wavelet transform.Firstly,we constructed filter bank by using construction method of 2-D non-separable wavelet filter banks applied to EZW algorithm.Then,we conduct a simulation experiment on our proposed method based on the 2-D non-separable wavelet transform and classical EZW algorithm.The experimental results show that,in the case of keeping the compression ratio slightly higher,the quality of reconstructed image in our proposed method is better and the PSNR is higher than that of the classical algorithm.So,the compression efficiency of the method proposed in this paper is better than that of the classical EZW algorithm.
Keywords/Search Tags:Image compression, 2-D non-separable wavelet transform, Filter bank, Embedded zero-tree wavelet encoding
PDF Full Text Request
Related items