We all take lousy, grainy, and jerky videos. And then we upload them to YouTube to torture our friends with the poor video quality.

But MotionDSP has a “video restoration” technology (see our coverage, which has demos) that can fix all that.

Dubbed FixMyMovie, it can take grainy and jerky video and turn it into smooth video with four times better resolution. Over time, the technology can consume a lot of graphics horsepower as it improves its ability to convert lousy videos into works of art. And that’s why Nvidia is announcing today that it has taken a stake in MotionDSP. The amount isn’t being disclosed.

Sean Varah, chief executive of the company, said that MotionDSP has a new version of its software, code-named Carmel, which uses sophisticated techniques to track objects across multiple frames of video. It then deletes the noise, or visual flaws, from the scene. And it can intelligently fill out the video with the right details if one of the frames becomes fuzzy or slows down. It does so by looking at about 60 frames of video and making comparisons among them. Then it mathematically enhances or deletes parts of the video, correcting for motion blur and poor lighting conditions.

The San Mateo, Calif.-based company debuted its FixMyMovie consumer technology at the DEMOfall 07 conference. It received a small amount of funding from the CIA’s In-Q-Tel venture fund last year and has previously deployed a high-end video-smoothing technology for security purposes. Varah said that the Secret Service, CNN, law enforcement agencies and others are using it in a “CSI”-style video forensics work. That version costs thousands of dollars, but the consumer version will debut at under $100 early next year.

The company has 17 employees and was founded in 2005. It raised its angel funds in March 2006 and has struggled to raise more money for a while now. Even after the funds from the CIA, its backing totaled a mere $1 million. So it is taking the money from Nvidia as part of its Series A (first round). Varah said the company continues to try to raise its first round, despite the difficult fundraising environment now.

The technology was originally initiated by Peyman Milansar, an electrical engineering professor at the University of California at Santa Cruz. Over the past few years, MotionDSP has adapted that work and refined it so it could be built into products.

Over time, Varah said the technology could be used to smooth out real-time video, such as video confererence calls. Varah said that MotionDSP is using Nvidia’s CUDA programming language to tap the graphics chip’s power for its real-time version.

By doing so, Varah said it can improve video performance five-fold. Nvidia is investing in the technology out of the belief that it can increase demand for its graphics chips if it can create non-gaming applications that can increase overall graphics processing demand.