Efficient Realization And Research Of Algorithms Of The 2-D Discrete Wavelet Transform

The two-dimensional (2-D) discrete wavelet transformation (DWT) is apowerful tool for digital image analysis. It is applied in many domains of image processing, such as image coding and compression, digital image processing, Nevertheless, 2-D DWT demands massive computations, and in applications of practical image processing requiring perfect reconstruction through inverse wavelet transformation. This needs undistorted boundary data. Thus, it is significant to realize 2-D DWT in image processing efficiently.In this paper, the basical conception of wavelet transformation is introduced at first. Then, the theory of multiresolution and 2-D tensor product multiresolution which are theoretical bases of discrete wavelet transformation are discussed and wavelet algorithms of analysis and reconstruction are strictly reasoned. The emphasis in this paper is exploration in realization of 2-D DWT in image processing. Through exploring the structure of nonseparable computation method applied to realize 2-D DWT using orthonormal wavelet base with periodical extension of image data and biorthonormal wavelet base with symmetrical periodical extension of image data and analyzing characters of 2-D wavelet filters which are produced by 1-D wavelet filters, and combining a new computational method, modified algorithms are proposed.Compared with the classical algorithms, modified algorithms need fewer numbers of multiplications and less precision for filters. Furthermore, modified algorithms are suitable for real-time application and realization because of similar structures of computing every sort of outputs.