MORPHEUS APPLICATIONS  TEST CASES AND VALIDATION

The whole MORPHEUS project is driven by the following set of applications:

DTO

Figure 1 depicts the DI workflow parts with the data ingest coming directly from the film scanner or from a film storage device. Although, the market share of digitally filmed material is constantly growing, the vast majority is still shot using conventional cameras requiring a later scanning of the film. The four typical processing steps of this first post-production phase are film grain noise reduction, secondary colour correction, scratch & dirt removal and video effects.

In a simplified view the major goal of these algorithms is to normalize the ingested material in order to become independent of film stock used, different light situation during acquisition and sub optimal results caused by the scanning process itself. All mentioned processing steps are usually manually adjusted to the scene being processed and employ fairly complex algorithms.

Today, most of these algorithms run offline on high performance standard PCs. In order to give the operator an immediate feedback of the settings chosen as well as to allow a direct ingest from the scanner without a slowdown by the subsequent processing steps, solutions for real-time implementation of these algorithms are under investigation. The preferred scenario is that one or more accelerator boards are plugged into a standard PC, which is already used to control the scanning or post-production process. All algorithms should be implemented to support real-time or near real-time processing, although it is not necessary to run all algorithms simultaneously as only a selection of them is typically employed on a particular scene.

Hence, in this context real-time means inline with the scanner output rate of up to 5 Gbit/s.

Figure 1 Structure of HD video/digital film processing unit

INTRACOM

The application targeted by Intracom Telecom is the emerging IEEE 802.16j standard. The latest standard currently in force from the IEEE 802.16 family is 802.16e, the basis for Mobile WiMAX technology. This standard mandates the use of Orthogonal Frequency Division Multiple Access (OFDMA) technology for the physical layer and provides all necessary support in the physical and MAC layers for mobility management, such as network entry, handover, etc. The next standard, 802.16j, currently in preparation, extends the concepts defined in .16e by adding the possibility of multi-hop communication between mobile and base station. For this, the Relay Station entity is defined. The relay station is connected to the base station on one side and to a group of mobile stations on the other. The connection to the base station, where the relay acts more or less as a subscriber/mobile station, is called the relay link, while the connection to the mobiles, where the relay acts as a simple base station, is called the access link. BS-MS communication now may take place over two hops (BS-RS and RS-MS), which can be advantageous because a poor channel is divided into two better ones, allowing for more spectrally efficient modulation and coding schemes. Alternatively, the range of a network cell can be extended, with relays placed near its periphery, serving distant mobiles. The 802.16j standard will reuse the OFDMA physical layer from .16e, with some minor enhancements possible, and will make significant amendments to the MAC layer.

Relays can be fixed (located e.g. on rooftops, lamp posts, etc), nomadic (transportable, eg on trucks) or mobile (on buses, trains, etc). Here we focus on the case of a Mobile Relay Station (MRS). Figure 2 demonstrates the concept of a multihop network, including an MRS mounted on a bus that provides service to passengers onboard. As the MRS moves within an area, it will have to perform handover between different base stations (when crossing from one network cell to another). At the same time the group of mobile stations it supports will also change dynamically over time. The physical layer mode used in each cell is determined by the base station that serves it. As the propagation environment differs from cell to cell (e.g. urban, suburban, rural), different base stations may require different physical layer modes. While simple terminals, supporting only the mandatory modes, are still backwards compatible with all base stations, they need to be able to support the advanced modes in order to take advantage of them. The same holds for a MRS, that acts as a terminal on the relay link.

Figure 2 ultihop network concept

ALCATEL-LUCENT

In the MORPHEUS project the contribution of Alcatel-Lucent mainly consists of the development of a concept which allows to update the hardware of installed network elements partially. This is achieved by a two step approach. In phase 1 of the project a demonstrator based on two FPGA evaluation boards is built in order to proof the concept. In phase 2 the application is ported to the MORPHEUS platform.

The application itself is described in the following:

Alcatel-Lucent is interested in being able to update particular hardware function in an ASIC architecture. The reason for such an update could be standard adaptations or bug fixes. The hardware functions which are likely to be updated must be selected carefully and allocated to an embedded FPGA. For demonstration purposes the well known Ethernet functionality has been chosen to be implemented. Figure 3 shows the basic concept. Ethernet data optionally contains reconfiguration data. This reconfiguration data is marked e.g. by using a specific IP address. Normal Ethernet data is routed along the system data path. Reconfiguration data, however, is collected packet by packet in a configuration RAM until it is complete. After a consistency check (CRC) is done successfully reconfiguration is executed under the control of the reconfiguration controller module.

Instead of the Ethernet protocol it is possible to use any other network protocol for the transmission of reconfiguration data. A patent for this concept is currently pending.

Figure 3 Reconfiguration of ASIC hardware via Ethernet protocol

TOSA

The system targeted by TOSA is a general purpose (multi applications) image processing / image understanding system for vision applications in the domain of surveillance. Such a system can be viewed as a large collection of real time algorithms which are activated (or not) in function of non predictable events such as the content of the image or an external information or a request from the user. It will be implemented through 3 levels of software as discussed in paragraph 5.2 of deliverable D5.3:

• A set of image processing operators which are implemented on the HREs of MORPHEUS and which provide most of the processing power
• A set of algorithms which run on a general purpose processor and which reconfigures and activates the HREs according to the set of operators needed by each algorithm
• A control programme which run on a general purpose processor and which selects and activates the algorithms in function of the situation

Figure 4 Three levels of software in intelligent surveillance system

Reconfiguration capability is a very important feature to implement such a system. This topic has been extensively discussed throughout paragraphs 5.1 to 5.5 of deliverable D5.1 and paragraph 5.2 of deliverable D5.2. In fact, 3 levels of reconfiguration are required:

• Real-time or in-process dynamic reconfiguration: capability to reconfigure into several different operators during the processing of an image so as to allow the implementation of complex image processing algorithms.
• Dynamic reconfiguration: capability to reconfigure into several different image processing algorithms so as to allow the activation of a given algorithm in function of a non predictable event such as the content of the image (intelligent behaviour) or an external information or a request from the user
• Static reconfiguration: capability to reconfigure into several different image processing algorithms so as to allow the use of the same device for a lot of different applications and a lot of different customers