COMPARISON OF LOSSY TO LOSSLESS COMPRESSION TECHNIQUES FOR DIGITAL CINEMA S. Andriani, G. Calvagno, T. Erseghe, G.A. Mian Dip. di Ingegneria dell’Informazione, Università degli Studi di Padova, Via Gradenigo 6/B, 35131 Padova, Italy M. Durigon, R. Rinaldo Dip. di Ingegneria Elettrica, Gestionale e Meccanica Università degli Studi di Udine Via delle Scienze 208, 33100 Udine, Italy M. Knee, P. Walland Snell & Wilcox Ltd. Southleigh Park House, Eastleigh Road Havant, Hants, PO9 2PE, United Kingdom M. Koppetz Arnold & Richter Cine Technik (ARRI) TFE Tuerkenstrasse 89 D-80799, Munich, Germany ABSTRACT This paper presents the investigation carried out for the MetaVi- sion project in the area of lossless compression for digital cin- ema. The goal of this investigation was to identify, among the many approaches, the one that is most suited to the MetaVision re- quirements. An extensive simulation campaign has thus identified JPEG-2000 as the best compromise between efficiency, standard- ization and lossy-to-lossless performance. This result was hardly predictable, especially because of the peculiar kind of material to be coded. 1. INTRODUCTION As time goes on, the transmission and recording bandwidth of available devices increases steadily. There is, however, a simi- lar pressure on the bandwidth required driven by the demand for resolutions which are closer and closer to a "film quality" ideal. Understandably, the creators of original material do not wish to see any compression performed which may cause artifacts to be created later in the production chain. The current "target" capture format for electronic film quality is 4k × 2k pixels, RGB 4:4:4 with a minimum 12 bits resolution at 24 frames per second. This implies a raw data rate of nearly 7 Gbit/s. In contrast the current standard rate for HD transmission of 1.48 Gbit/s represents the transmission rate available for real- time transfer and storage onto disk or tape. For good motion portrayal there is a need to increase the cap- tured frame rate (72 frames per second as a minimum, 150 frames per second preferred) and there is already talk of IMAX resolutions being captured at 8k × 4k pixels. It is fair to say that electronic sensors and cameras which can give this level of performance are not yet available, however it is equally clear that as technology advances on this front so the requirement to transfer and record higher and higher data rates will always be in excess of the avail- able transfer rates. The MetaVision project, funded by the EU IST programme, has been investigating ways to capture and compress images at electronic film quality, and how to use it to assist the post-producti- on workflow [1]. Now, in its final year, the project has developed MetaVision is the IST-20859 project supported by the European Com- mission in the context of Framework Programme 5. MetaVision started in October 2000 and completed successfully on 31st December 2003. http://www.ist-metavision.com/. Fig. 1. The MetaVision camera demonstrator. a camera system demonstrator [2] capable of capturing 2k × 1k pixels RGB images with 10 bits resolution running at 72 frames per second. With respect to the compression aspects of the MetaVision system, mathematically lossless or near lossless compression has being adopted in order to provide a reduction in bit rate. Although lossy compression for video has now reached quite a mature stage, the application of lossless compression to video has not received much attention in the past because of the limited compression ra- tios available. However, with this emerging need for very high quality handling of content in production and post-production it is necessary to develop techniques which will work in this environ- ment. In this paper we report the main results of the research on com- pression algorithms carried out in the context of MetaVision (e.g., see [3], [4], [5]). The paper is organized as follows. In Section 2 we describe the MetaVision camera demonstrator and in Section 3 we clarify what compression is needed in MetaVision. In Section 4 we present the most interesting lossless methods developed in this research and show their performances in Section 5. Section 6 re- ports some conclusions.