GPU: Application for CCTV systems

Current generation GPU´s are really powerful, with lots of stream processors, fast memory and in some cases hardware video decoding support. Their prices are low and support multiple high resolution outputs. Its main use is in CAD applications and Video-games due they geometry processing capabilities, but they are really good at image processing. That makes them ideal for CCTV rendering. Traditional CCTV systems are based on a stack of decoder cards with analog outputs. These systems are expensive and hard to install due to its physical connections. We can replace such a system with a PC with a good GPU card. This system presents some advantages over the analog solution: it can use an existing network installation, can show any source that the PC can receive (HTML, VNC, Capturers, etc), can mix analog and IP cameras and the GPU shader capabilities permits apply complex filters to the video. Moreover, we can use the GPU video analysis (motion detection, face recognition, etc). In this paper we present a system aid to decode and render multiple video sources, based on a PC platform (but feasible in any platform with a modern GPU, we have it running on a Raspberry Pi) capable of decoding and rendering close to 100 video streams on a current system, with support of 24 monitors in a single machine. This system is already working and installed in multiple places with good results.