Showing papers by "Gemma Piella published in 2005"
TL;DR: A technique for building adaptive wavelets by means of an extension of the lifting scheme that comprises an adaptive update lifting step and a fixed prediction lifting step that can choose between two different update filters.
48 citations
TL;DR: This paper presents the construction of adaptive wavelets by means of an extension of the lifting scheme and shows that these adaptive schemes yield lower entropies than schemes with fixed update filters, a property that is highly relevant in the context of compression.
Abstract: Over the past few years, wavelets have become extremely popular in signal and image processing applications. The classical linear wavelet transform, however, performs a homogeneous smoothing of the signal contents which, in some cases, is not desirable. This has led to a growing interest in (nonlinear) wavelet representations that can preserve discontinuities, such as transitions and edges. In this paper, we present the construction of adaptive wavelets by means of an extension of the lifting scheme. The basic idea is to choose the update filters according to some decision criterion which depends on the local characteristics of the input signal. We show that these adaptive schemes yield lower entropies than schemes with fixed update filters, a property that is highly relevant in the context of compression . Moreover, we analyze the effect of a scalar uniform quantization and the stability in such adaptive wavelet decompositions.
24 citations
14 Nov 2005
TL;DR: A new class of adaptive wavelet decompositions that can capture the directional nature of picture information and establish the conditions under which these decisions can be recovered at synthesis, without the need for transmitting overhead information.
Abstract: We present a new class of adaptive wavelet decompositions that can capture the directional nature of picture information. Our method exploits the properties of seminorms to build lifting structures able to choose between different update filters, the choice being triggered by a local gradient of the input. In order to discriminate between different geometrical information, the system makes use of multiple criteria, giving rise to multiple choice of update filters. We establish the conditions under which these decisions can be recovered at synthesis, without the need for transmitting overhead information.
20 citations
Proceedings Article•
01 Sep 2005
TL;DR: A multiresolution approach based on a special class of adaptive wavelets which allows the extraction of salient features without loosing accuracy is proposed.
Abstract: In this paper we propose a feature-based wavelet representation for image classification and visualization The work is primarily motivated by the need to classify quickly and efficiently large multispectral satellite images, and possibly to perform the classification task directly on compressed data We propose a multiresolution approach based on a special class of adaptive wavelets which allows the extraction of salient features without loosing accuracy