Conference Paper

Spatially Continuous Orientation Adaptive Discrete Packet Wavelet Decomposition for Image Compression

New South Wales Univ., Sydney, NSW
DOI: 10.1109/ICIP.2006.312613 Conference: Image Processing, 2006 IEEE International Conference on
Source: IEEE Xplore


In this paper, we propose an orientation adaptive discrete wavelet transform (DWT) with perfect reconstruction. The proposed transform utilizes the lifting structure to effectively orient the 2D-DWT bases in the direction of local image features. A shifting operator is employed within each lifting step to align spatial geometric features along the vertical or horizontal directions. The proposed oriented transform generates a scalable representation for the image and the orientation information. To approximate the asymptotically optimal rate-distortion performance of a piecewise regular function more closely, we adopt a packet wavelet decomposition. The experimental results obtained by implementing the proposed transform in a JPEG2000 codec illustrate superior compression performance for the oriented transform with more than 2.5 dB improvement for highly oriented natural images. More importantly, even at the same PSNR, the proposed scheme reduces the visual appearance of the Gibbs-like artifacts significantly, considerably improving the visual quality of the reconstructed image

Full-text preview

Available from:
  • Source
    • "The principal disadvantage of the LS described above, is that the linear filtering structure is fixed and thus, cannot match well the sharp transitions in the signal. The lifting schemes with adaptive prediction (APLS) [10], [11], [12] or adaptive update (AULS) [7], [8], [9] have been designed to overcome this limitation by the use of a filter that is able to adapt itself to the input signal it is analyzing. In the following, we will focus on the APLS and describe briefly its principle. "
    [Show abstract] [Hide abstract]
    ABSTRACT: In this paper, we study the problem of image denoising by using an adaptive lifting scheme. Such a scheme can adapt itself well to the analyzed signal, which allows to keep important information for denoising applications. However, it results in a non-isometric transform which can be an important limitation as most of the denoising approaches rely on the estimation of the noise energy in the subbands. In a previous work, it has been shown how to evaluate the subband energies of an uncorrelated signal, in the wavelet domain when using such an adaptive scheme. Based on this previous work, we propose in this paper an estimation of the noise energies in the subband and use it to perform image denoising. Experimental results illustrate that this approach is more effective, in image denoising, than the classical non adaptive lifting schemes both considering perceptual and non perceptual image quality measures.
    Full-text · Conference Paper · Jul 2010
  • Source
    • "The great flexibility of lifting scheme offers the possibility to replace linear filters by nonlinear ones. In particular, LS with adaptive update [2], [3], [4] or adaptive prediction [5], [6], [7] have been proposed in the literature, with the target of avoiding oversmoothing of important features such as borders, and at the same time of exploiting the correlation of homogeneous regions by using long filter on them: different filters are thus used in different parts of image, and so the entire transform can be strongly non-isometric. This can be a serious obstacle, since all most successful coding techniques rely on the distortion estimation in the transform domain, either explicitly like in the EBCOT algorithm [8], or implicitly, like in zero-tree based techniques like SPIHT [9] and EZW [10]. "
    [Show abstract] [Hide abstract]
    ABSTRACT: The lifting scheme represents an easy way of implementing the wavelet transform and of constructing new content-adapted transforms. However, the adaptive version of lifting schemes can result in strongly non-isometric transforms. This can be a major limitation, since all most successful coding techniques rely on the distortion estimation in the transform domain. In this paper we focus on the problem of evaluating the reconstruction distortion (due to quantization noise) in the wavelet domain when a non-isometric adaptive-prediction lifting scheme is used. The problem arises since these transforms are nonlinear, and so common techniques for distortion evaluation cannot be used in this case. We circumvent the difficulty by computing an equivalent time-varying linear filter, for which it is possible to generalize the distortion computation technique. In addition to the theoretical formulation of the distortion estimation, in this paper we provide experimental results proving the reliability of this estimation, and the consequent improvement of RD performance, thanks to a more effective resource allocation which can be performed in the transform domain.
    Full-text · Conference Paper · Nov 2009
  • Source
    • "As correlation persists, large magnitude coefficients representing contours remain spatially colocated . As a result, current research efforts are focusing on the design of schemes able to decorrelate these coefficients or to minimize their energy [1] [2] [3] [4] [5] [11]. "
    [Show abstract] [Hide abstract]
    ABSTRACT: In this paper we introduce an adaptive local pdf estimation strategy for the construction of Generalized Lifting (GL) mappings in the wavelet domain. Our approach consists in trying to estimate the local pdf of the wavelet coefficients conditioned to a context formed by neighboring coefficients. To this end, we search in a small causal window for similar contexts. This strategy is independent of the wavelet filters used to transform the image. Experimental results exhibit interesting gains in terms of energy reduction comparable to those obtained in [8]. In order to take benefit from this energy reduction, specific entropy encoder should be designed in the future.
    Full-text · Conference Paper · Jun 2009
Show more