|
 
|
|
|
|
|
|
|
|
|
Roberto
Rinaldo
professor
via
delle Scienze, 208
33808 - Udine
tel. +390432558288
fax +390432558251
rinaldo@uniud.it
office hours: Wednesday 11-12
 pagine in italiano
Extended curriculum (Italian and English)
|
|
Roberto Rinaldo obtained the "Laurea in Ingegneria
Elettronica" degree in 1987 from the University of Padova, Padova,
Italy. From 1990 to 1992, he was with the University of California at
Berkeley, where he received the MS degree in 1992. He received the Doctorate
degree in "Ingegneria Elettronica e dell'Informazione" from
the University of Padova in 1992. In 1992 he joined the Dipartimento
di Elettronica e Informatica of the University of Padova as a "ricercatore".
Starting from November 1st 1998, he was associate professor in Communications
and Signal Processing in the same Department.
Since November 2001, he was an associate professor in the "Dipartimento
di Ingegneria Elettrica, Gestionale e Meccanica" of the University
of Udine. Starting December 2003, he is now a professor in the same
department. His interests are in the field of multidimensional signal
processing, video signal coding, fractal theory and image coding. Prof.
RInaldo is part of the Telecommunication and Signal Processing group
(TSP) of the University of Udine..
|
Roberto Rinaldo's research activity is in the
general field of multidimensional signal processing and analysis, with a
particular interest in image and video coding in telecommunication systems.
He is now considering the problem of digital transmission of multimedia
signals in telecommunication networks.
A list of the main research topics he is working
on or worked on in the past is the following:
Object based video coding, composite
document coding and lossless coding
The standard
approach to video compression consists in computing a motion-compensated
prediction of the actual frame image, followed by transform coding of the
residual. More recently, object oriented video processing has received
increasing attention and in fact the most recent standards describe the
scene as a collection of objects that can be animated and coded
independently. In this project, we considered the problem of coding and
also of frame interpolation in reduced rate video transmission systems. The
objects in the scene are modelled as the projection onto the camera plane
of 3D rigid bodies undergoing three dimensional motion. A Kalman filter is
used to estimate motion parameters on the basis of the observations
relative to the 2D positions of characteristic features in the scene. The
estimated motion is used to interpolate missing frames at the receiver. Another
application is coding, where the residual error between original and motion
compensated frames is actually sent to the receiver. More recently, a
similar scheme has been proposed for the animation of virtual faces in a
model based video coder. Local motion relative to the mouth and eye
movements is also considered and estimated using a template matching
procedure.
Coding of composite documents requires, for the regions corresponding to
text or graphic art, the use of lossless coding techniques. The schemes
proposed for this application use spatially variant prediction and entropy
coding of the residual. Recently, the JPEG2000 standard for still images,
provides a scheme for progressive lossy to lossless coding of images. We
are now considering, within the European project "MetaVision",
the challenging problem of lossless compression of colour video with cinema
quality, starting from a raw bit rate of more than 3 Gbit/s.
back
See also:
G. Calvagno, F. Fantozzi and R. Rinaldo, "Feature based global and local
motion estimation for videoconference sequences", Proceedings 2001 International
Conference on Image Processing (ICIP 2001), Tessaloniki, Greece, October 7-10,
2001.
G. Calvagno, G.A. Mian, R. Rinaldo, "3D Motion Estimation for Frame Interpolation
and Video Coding," Proc. Int. Workshop on Packet Video, Cagliari, May 2000.
G. Calvagno,
G.A. Mian, R. Rinaldo, L. Zimolo, "Ancient Document Compression and Archiving
Via Page Decomposition", Proc. EUSIPCO 2000, Tampere, Sept. 2000.
G. Calvagno, R. Rinaldo and L. Sbaiz, "3-D Motion Estimation of Objects
for Video Coding," IEEE Journal on Selected Areas in Communications, Vol.
16, No. 1, pp.86-97, Jan. 1998.
G. Calvagno, V. Orsatti, L. Sbaiz and R. Rinaldo, "Object Oriented Coding
Using 3D Motion Estimation", Proceedings of IX European Signal Processing
Conference (EUSIPCO'98), Rhodes (Greece), September 8-11, 1998, pp. 2085-2088.
Image and Video coding
The enormous amount of information required for the representation,
transmission or storage of multimedia signals, in particular for video and
images, requires the adoption of sophisticated source coding techniques in
the first place. Research in this field focused on the theory of subband
decomposition, and in particular of pyramidal schemes based on the wavelet
representation. The wavelet transform opearates a multiresolution
representation, which exhibits a certain amount of redundancy which can be
exploited for efficient compression. In particular, quantized zero
coefficient positions appear to be highly correlated across scales.
Moreover, a certain degree of similarity appears around edges. An image
coder which exploits redundancy of the wavelet representation was developed
for this project. Similarly to what is done in fractal coders, blocks of
coefficients in a subband are predicted from coefficients in lower
resolution subbands with the same orientation. Scalar quantization followed
by entropy coding or pyramid vector quantization is used for those
coefficient blocks which could not be effectively predicted.
Similar ideas were used for the design of a video coder for the
transmission of videoconference sequences at low bit rates (< 64
kbit/s). It exploits the redundancy of the wavelet representation both in
the 2D and 3D domain. In particular, a segmentation mask is used to
discriminate wavelet coeefficients which could not be effectively predicted
using motion compensation, and which are coded using the prediction scheme
outlined above in the 2D motion compensated prediction error image.
back
See also:
G. Bologna, G. Calvagno, G.A. Mian, and R. Rinaldo, "Wavelet Packets and
Spatial Adaptive Intraband Coding of Images", Signal Processing: Image
Communication, Vol. 15, No. 10, August 2000, pp. 891-906.
G. Calvagno, R. Rinaldo, "Multiresolution Vector Quantization for Video
Coding," Multidimensional Systems and Signal Processing, Vol. 8, No. 1,
Jan. 1997, pp. 129-150.
R. Rinaldo, G. Calvagno, "Hybrid Vector Quantization for Multiresolution
Image Coding," IEEE Transactions on Image Processing, Vol. 6, No. 5, May
1997, pp. 753-758.
R. Rinaldo, G. Calvagno, "Image Coding by Block Prediction of Multiresolution
Subimages," IEEE Trans. on Image Processing, vol. 4, no. 7, pp. 909-920,
July 1995.
R. Rinaldo, G. Calvagno, "An Improved Wavelet-Fractal Coder," in Proc.
ISCAS 1994, pp. 3.113-3.316, London, May 1994.
R. Rinaldo, G. Calvagno, "Image Coding by Block Prediction of Wavelet Coefficients,"
in Proc. EUSIPCO 1994, Edinburgh, September 1994, pp. 1249-1252.
R. Rinaldo, G. Calvagno, "An Image Coding Scheme Using Block Prediction
of the Pyramid Subband Decomposition," in Proc. IEEE ICIP 1994, Austin
TX, November 1994, pp. 878--881.
G. Calvagno, G.A. Mian, R. Rinaldo, "Coding Gain of Intra/Inter-Frame Subband
Systems," Proc. 1995 Data Compression Conference (DCC 95), Snowbird, Utah,
March 1995.
G. Calvagno, R. Rinaldo, "Low Bit Rate Video Coding Based on Generalized
Prediction of the Pyramid Subband Decomposition," IEEE 1995 International
Conference on Image Proc., vol. II, pp. 422-425, October 1995.
Statistical modeling of signals in
the subband decomposition
The signals in a subband decomposition of images can be modelled
approximately as independent vectors of independent Laplacian or
generalized Gaussian random variables. Using this model, it is possible to
evaluate the entropy of the output symbols and the distortion resulting
from the use of a given quantizer. In this project, we designed a procedure
to determine the optimal threshold quantizer for subband coding. This
quantizer has a dead-zone around the origin and is used in various
standards for video coding. The availability of a procedure to design the
optimal quantizer and compute the resulting rate-distortion curve allows
for a precise bit-rate or distortion control in video transmission systems.
A better model for images is a first order Markov process. Using this
model, it is possible to compute the covariance matrix of subband vectors,
and not just use the independence assumption. We developed an efficient
procedure in the time domain to compute the coding gain in subband
decomposition systems, both for intraframe and inter-frame coding. We also
considered the problem of optimal reconstruction filter design and of wavelet
best basis using this model.
back
Multiresolution analysis and wavelet
transform for fractal images
Research in this field has been carried out at the Department of Electrical
Engineering of the University of California at Berkeley, in the years
1990-1992. Fractal geometry is a useful tool to build deterministic and
stochastic models for many natural phenomena. As a consequence, fractal
theory has been used in a variety of problems, including image segmentation
and coding, or generation of virtual worlds in computers graphics. From a
general stand-point, fractal images exhibit a high degree of
self-similarity when analyzed at different scales. In other words, a
fractal image, when analyzed through a magnifying lens, is likely to
exhibit the same properties and characteristics of the original image. The
wavelet transform, which provides a multiresolution analysis of signals, is
therefore a promising tool for fractal image analysis. Moreover, the
self-similarity property of fractals suggests that a multiresolution
representation can provide a very efficient method for coding or
representation. In this project, we addressed the "inverse
problem" of determining the parameters of a class of fractal objects,
generated using iterated maps (IFS), which could approximate a given image.
The solution is obtained using the wavelet transform and the moment method.
back
See also:
R. Rinaldo, A. Zakhor, "Fractal Approximation of Images," in "Control
and Dynamic Systems", vol. 67, C. T. Leondes Ed., Academic Press, pp. 91-141,
1994.
R. Rinaldo, A. Zakhor, "Inverse and Approximation Problem for Two-Dimensional
Fractal Sets," IEEE Trans. on Image Processing, vol.3, (no.6), pp. 802-20,
Nov. 1994.
R. Rinaldo, D. Taubman and A. Zakhor, "Applications of Multi-Resolution
Analysis to Images," Seventh Workshop on Multidimensional Signal Processing.
September 1991, Lake Placid, New York.
R. Rinaldo, A. Zakhor, "Inverse Problem for Two-dimensional Fractal Sets
Using the Wavelet Transform and the Moment Method," Proc. 1992 IEEE ICASSP,
March 1992, S. Francisco, IV-665.
R. Rinaldo, A. Zakhor, "Fractal Approximation of Images," Data Compression
Conference, 1993, Snowbird, USA.
R. Rinaldo, A. Zakhor, "Inverse Problem and Approximation of Fractal-like
Images,'" in. ISCAS 1994, pp. 3.109-3.312, London, May 1994.
Spectral estimation
of television sequences
Quantitative information on frequency characteristics of signals is very
useful for filter design and coding. In this project, a tool for the
spectral estimation of television signals was developed, based on the
multidimensional Periodogram and autoregressive models. The extension to
the multidimensional case of the techniques usually considered for 1D
signals is quite challenging, due to the enormous amount of data required
for estimation.
The spectrum analyzer was used for the design of 3D filters for luminance
and chrominance separation in the PAL signal. The filter design was carried
out by taking into account the 3D frequency characteristics of the signal
and using an MSE criterion. Applications to High Definition Television and
the problem of algorithm optimization were also considered.
back
See also:
G. Cortelazzo, G. A. Mian, R. Rinaldo, "Toeplitz properties of the block
matrices encountered in the processing of spatio-temporal signals", IEEE
Trans. Signal Process., Vol. 39, (no.7), pp. 1672-4, July 1991.
G. M. Cortelazzo, G. A. Mian, R. Rinaldo, "Spectral estimation of video
signals," Multidimensional Systems and Signal Processing, V3 N2-3, pp.
131-160, May 1992.
R. Rinaldo, G. M. Cortelazzo, G. A. Mian, "Techniques for the efficient
evaluation of multidimensional autocorrelation functions", IEEE Trans.
on Signal Process., vol. 40, no. 11, pp. 2854-7, Nov. 1992.
G. Cariolaro, R. Rinaldo, L. Tomba, "A bidimensional Model of Line-Shuffling",
Image Communication, pp. 291-304, December 1990. G. Cariolaro, F. Molo, R. Rinaldo,
"Spectral Analysis in HD-MAC Systems", Third International Workshop
on HDTV, 30 Aug-1 Sept. 1989, Torino.
G. Cortelazzo, G. A. Mian, R. Rinaldo, "On multidimensional FIR filters
optimal for luminance/chrominance separation", 1990 IEEE ISCAS, May 1990,
New Orleans (USA).
R. Rinaldo, G. M. Cortelazzo, R. Bernardini, "Extension of Rader's algorithm
for High Speed Multidimensional Autocorrelation", Proc. SPIE - Image Processing,
November 1991.
R. Rinaldo, G. M. Cortelazzo, G. A. Mian, "An efficient algorithm for the
evaluation of Multidimensional Autocorrelation functions", 1991 IEEE ICASSP,
Toronto, Canada.
2003 European
Project "WireNet", RTD Performers: Sintef (NO), Labor (I)
1999 "Analisi, sintesi e indicizzazione di sequenze audio/video
nella comunicazione multimediale", Coordinator: Prof. Carlo Braccini
(Univ. Genova), Financing: MURST (40%)
1999 European Project IST-1999-20859 "Metavision", Partecipanti:
Snell & Wilcox (UK), BBC (UK), INESC (P), Electrocraft (UK), France 2
(F), ARRI (D), MRT (N), Università di Padova (I)
1998 "Ambienti avanzati per comunicazioni a pacchetto," Coordinator
Prof. Enrico Gregori (CNR CNUCE Pisa), Financing: CNR
1997 "Comunicazione
multimediale: analisi di contenuto, modellamento, accesso e protezione dell'informazione",
Coordinator Prof. Carlo Braccini (Univ. Genova), Financing: MURST (40%)
1995 "Elaborazione
e codifica di segnali per sistemi multimediali di telecomunicazione",
Coordinator Prof. Carlo Braccini (Univ. Genova), Financing: MURST (40%)
1996 "Elaborazione e codifica di segnali per sistemi multimediali
di telecomunicazione", Coordinator Prof. Carlo Braccini (Univ. Genova),
Financing: MURST (40%)
1997 "Comunicazione
video in ambito radio-mobile", Coordinator Prof. Riccardo Leonardi (Univ.
Brescia), Financing: CNR
1997 "Ambienti
avanzati per comunicazioni a pacchetto", Coordinatore: Prof. Enrico Gregori
(CNR CNUCE Pisa), Financing: CNR
1996 "Progetto
di un codificatore video a basso bit rate", Coordinatore: Prof. Riccardo
Leonardi (Univ. Brescia), Financing: CNR
1995 "Progettazione
di un sistema di comunicazione video a bassissimo bit-rate", Coordinator
Prof. Riccardo Leonardi (Univ. Brescia), Financing: CNR
back
He teaches the courses:
Telecommunications Networks, Electrical Communications
and Telecommunication Systems II.
For a brief course description and class material (in Italian), please click
here.
See also:
G. Calvagno, G.A. Mian, R. Rinaldo, W. Trabucco, "Two-Dimensional Separable
filters for Optimal Reconstruction of JPEG-Coded Images," IEEE Transactions
on Circuits and Systems for Video Technology, Vol. 11, No. 7, July 2001, pp.
777-787.
G. Calvagno, G.A. Mian and R. Rinaldo, "Synthesis Filter Bank Optimization
in Two Dimensional Separable Subband Coding Systems", IEEE Transactions
on Image Processing, Vol. 9, No. 9, September 2000, pp. 1497-1508.
G. Calvagno, G.A. Mian and R. Rinaldo, "JPEG Decoding via Two-Dimensional
Optimal Separable Filters", Proceedings 2001 International Conference on
Image Processing (ICIP 2001), Tessaloniki, Greece, October 7-10, 2001.
G. Calvagno, M. Cantagallo, G.A. Mian and R. Rinaldo, "Synthesis Filter
Bank Optimization in 1D and 2D Separable Subband Coding", in Multimedia
Communications, F. De Natale and S. Pupolin (Eds). London: Springer-Verlag, 1999, pp. 18-31.
G. Calvagno, C. Ghirardi, G.A. Mian
and R. Rinaldo, "Modeling of Subband Image Data for Buffer Control,"
IEEE Transactions on Circuits and Systems for Video Technology, Vol. 7, No.
2, April 1997, pp. 402-408.
G. Calvagno, G.A. Mian, R. Rinaldo, "On the Computation of the Coding Gain
for Subband Coders," IEEE Trans. on Communications, vol. 44, n. 4, pp.
475--487, April 1996.
G. Calvagno, C. Ghirardi, G.A. Mian and R. Rinaldo, "A buffer control technique
for video coding", Proc. 1995 European Symposium on Advanced Networks and
Services: Compression Technologies and Standards for Image and Video Communications,
Amsterdam, March 1995, pp. 31-40.
G. Calvagno, G.A. Mian, R. Rinaldo, "Coding Gain of Intra/Inter-Frame Subband
Systems," Proc. 1995 Data Compression Conference (DCC 95), Snowbird, Utah,
March 1995.
Bilato-S, Calvagno-G, Mian-GA, Rinaldo-R, "Accurate bit-rate and quality
control for the MPEG video coder" Proceedings. International Conference
on Image Processing, Santa Barbara, CA, USA; 1997; 3 vol., pp. p.571-4 vol.3.
Multiple Description
image and Video Coding
Robust transmission of multimedia
streams over error prone networks has recently become an important issue for
many applications. Typical scenarios include wireless communication and distributed
networking. The transmission of multimedia signals on such networks can make
unpractical the usual mechanism of acknowledge and packet retransmission, because
of application sensitiveness to delays. Moreover, a return channel may not be
available or inconvenient to use. The use of error correction techniques may
also not help when packets are actually dropped or when one of the different
paths used to send the bit stream, as in layered coding, totally fails. Multiple
Description coding is a recently proposed solution, where several description
of the source are sent over independent paths. The idea is to send a redundant
description of a single source to the receiver and be able to reconstruct the
transmitted data when part of this information is lost. Among the different
MD techniques known in the literature we are particularly interested in the
use of redundant bases
(frames). Frame expansions permit exact reconstruction of lost information
even in the presence of data losses, at the expense of added redundancy. In
this sense, the technique is similar to MD Forward Error Correction coding,
a recently proposed scheme where redundancy is added across packets of different
descriptions of the source. The possible advantage in using frames is that added
information can be visually consistent, so that every description contains information
that can help to reconstruct or approximate the original sequence.
See also:
R.
Bernardini and R. Rinaldo, "Efficient
reconstruction from frame based multiple descriptions,'' to appear in IEEE
Trans. Signal Proc.
R. Bernardini and R. Rinaldo, "Frame-based
multiple description video coding with extended orthogonal filterbanks,"
to appear in EURASIP Journal of Applied Signal Processing.
A. Vitali, F. Rovati, R. Rinaldo, R. Bernardini, and M. Durigon, "Video
streaming over lossy/variable bandwidth networks by means of multiple description,"
in Proceedings of MMSP 2004, (Siena), pp. 498--501, IEEE, Sept. 2004.
R. Bernardini, R. Rinaldo, A. Tonello, and A. Vitali, "Frame based multiple
description for multimedia transmission over wireless networks," in Proceedings
of WPMC 2004, July 2004.
R. Bernardini and R. Rinaldo, "A robust iterative algorithm for reconstruction
from redundant filter banks," in Proceedings of ICASSP 2004, IEEE, May
2004.
R. Bernardini, M. Durigon, R. Rinaldo, L. Celetto, and A. Vitali, "Polyphase
spatial subsampling multiple description coding of video streams with h264,"
in Proceedings of ICIP 2004, (Singapore), pp. 3213--3216, Oct. 2004.
R. Bernardini, M. Durigon, and R. Rinaldo, "Low-delay reconstruction of
punctured frame-coded streams," in Asilomar Conference on Signals, Systems,
and Computers, (Pacific Grove, CA), Nov. 2003.
>