Maria Grazia Albanesi

Image compression by the wavelet decomposition

Decomposition of images with the Haar orthonormal basis which is an important member of compactly supported Wavelets and a quadtree structured hierarchical coding technique are used in this work to obtain high image compression efficiency and time complexity linear in the number of pixels. An exhaustive testing of the algorithm has been performed on images of different complexity and typical of some application environment (image transmission and storing, remote control of intelligent robots). The results of the experiments arc presented and discussed. Finally, a comparison of quality performance of these techniques with the JPEG (Block Cosine Transform coding) compression technique is presented.

Quantitative assessment of qualitative color perception in image database retrieval

We propose a multiresolution indexing algorithm based on color histogram which exploits the wavelet decomposition and a customized quantization for content-based image retrieval. The aim is to extract automatically the chromatic content of the images and to represent it with simple, robust, efficient and low computational cost descriptors. The proposed method has been integrated for a complete CBIR system, where the classification of images is performed on a qualitative subjective color perception. The system allows testing the semantic and chromatic class homogeneity previously defined by a human observer. Experimental results have been evaluated by the quantitative assessment parameters (averaged precision and recall). Multiresolution proved to be a valid framework to introduce spatiality in color histogram indexing, to dramatically decrease the computational complexity and to validate the qualitative subjective classification.

Wavelets and human visual perception in image compression

The paper describes a new adaptive coding technique for still, grey-level images. The algorithm is based on a multi-resolution decomposition on biorthogonal wavelet basis and it includes different nonlinear models of the human visual perception in the compression task. Quantization is adaptive, and it is based on considerations about the contrast experiments which aim at determining in which conditions (contrast of the input image, compression ratio) the inclusion of the models improves the performance. The results are compared with standard techniques which do not include considerations about human visual perception.

An HVS-based adaptive coder for perceptually lossy image compression

Abstract In this paper a locally adaptive wavelet image coder is presented. It is based on an embedded human visual system model that exploits the space- and frequency-localization properties of wavelet decompositions for tuning the quantization step for each discrete wavelet transforms coefficient, according to the local properties of the image. A coarser quantization is performed in the areas of the image where the visibility of errors is reduced, thus decreasing the total bit rate without affecting the resulting visual quality. The size of the quantization step for each DWT coefficient is computed by taking into account the multiresolution structure of wavelet decompositions, so that there is no need for any side information to be sent to the decoder or for prediction mechanisms. Perceptually lossless as well as perceptually lossy compression is supported: the desired visual quality of the compressed image is set by means of a quality factor. Moreover, the technique for tuning the target visual quality allows the user to define arbitrarily shaped regions of interest and to set for each one a different quality factor.

A taxonomy for image authentication techniques and its application to the current state of the art

In this paper we propose a taxonomy for image authentication techniques that takes into account three different features: the level of integrity verification, the approach to the generation of the authenticator and the capability of localizing manipulated areas. The goal is to revise the current and very heterogeneous bibliography on this topic, and to define a set of basic requirements for an authentication service, in order to identify which ones have already been met and which ones remain challenging. The major algorithms proposed in the last decade are examined according to the classification criteria; advantages and drawbacks of the possible approaches are reported. In particular, we investigate the relationship between the well established digital signature techniques and the emerging watermarking approaches. As a result, we propose a methodological approach to the design of robust, secure and efficient authentication algorithms.

Benchmarking Hough Transform Architectures for Real-Time

This paper reviews the Hough transform hardware implementations, with a specific analysis of the architectures that explicitly address the “real-time” issue. The work presents an introduction for a critical assessment of the notion of “real-time”, especially for what concerns modern multimedia applications. The main contribution of this work is the proposal of a new metric for measuring the performance of Hough transform architectures against a given definition of “real-time”. The basic idea is that there is no single set of constraints that define “real-time” for every application domain, and that even the simplest case of Hough transform for line detection, must be properly characterized within a specific application domain. The architectures are classified and evaluated, after a proper characterization of the Hough transform complexity, in terms of dimensions of parameter space and time complexity.

Shape detection with limited memory

Abstract A new approach to shape detection through the generalized Hough transform is introduced. The method is based on a limited memory implementation of the transform, that reduces its cost and makes it suitable for hardware implementation. The rationale of the method is that a shape is bound by a circle whose radius is, in most practical situations, much smaller than the dimensions of the image processed. This a priori knowledge can be used during the vote collection phase of the transform to guide flushing operations against a filled memory. The method is tested in the simple case of circles detection and in more practical situations of IC inspection.

A compact wavelet index for retrieval in image database

In this paper, we present an algorithm to build an index for image retrieval that heavily exploits the wavelet multi-resolution decomposition. The index can be constructed using alternative levels in the multi-resolution decomposition; it is based on the correlation between the scaled-down approximation image and its reconstruction obtained using only one level in the hierarchy of detail images. The method has been tested on RBG and YUV color spaces on a database of 167 images with good results. The effect of white Gaussian noise is negligible, because the wavelet approximation smooths off this kind of noise. The method is more sensitive to geometric distortion noise.

A HIGH SPEED HAAR TRANSFORM IMPLEMENTATION

This paper presents an implementation of the Haar transform suitable for VLSI integration. It shows how to map a bidimension linear transformation, which has a straightforward multiresolution realization on a pyramid data-parallel computer, onto a pipeline of simple processors. A further simplification of the linear structure leads to an extremely simple implementation based on a two-stage pipeline, capable of processing images as large as 1024×1024 pixels. VLSI simulations with current technologies predict HDTV video rates. Data compression is among the applications that benefit from the new formulation of the transform.

Robust Hierarchical Indexing based on Texture Features

In this paper, we present a hierarchical indexing method based on texture characterization for image retrieval. The novelty of our contribution is the hierarchical structure of the index: it exploits the multiresolution formulation of Wavelet Transforms to define a new set of approximated versions of the images for each level of resolution. On this set, the algorithm extracts significant signatures by means of statistical correlations; the experimental results and the analysis of computational complexity have proved that the algorithm presents the best performance at the highest level of the indexing hierarchy, where the computational complexity is the lowest. Our method has been evaluated by the following methodologies: (a) the study of the computational complexity for signature generation; (b) the comparison with analogous methods based on texture analysis by reporting the performance obtained on the same database (Brodatz); and (c) the evaluation of the robustness of the hierarchical indexing in different color spaces, by querying the database with different versions of the original images obtained by noise addition (gaussian and scanner acquisition noise and lossy compression distortion), brightness and contrast enhancement, color and scale adjustment and rotation. Even if our method is designed for texture databases, experiments show satisfactory results also on a real heterogeneous photographic database. This confirms the possibility of exploiting our method as a low computational complexity indexing tool based on texture characterization in a broader system for hierarchical content-based retrieval.