"Video is going mainstream, but there are issues hampering its quality in the state of the raw video and the computation and bandwidth limits of the systems handling it," said Ian Xie, founder and chief executive of Ipera.

Previously, Xie founded startup Mediostream which developed a real-time software encoder for MPEG-2. It was later licensed to Sony and others.

The idea Xie and a handful of other engineers developed was a real-time algorithm that could make a range of improvements in the quality of a video stream without requiring additional bandwidth or extensive computation power. The company's Pixel Magic software analyzes one or more frames of video at a time and applies pixel-level enhancements at 30 frames/second.

"Measures of video quality are very subjective, but we make obvious improvements in color saturation, contrast, skin tones and noise levels," said Xie. "We offer more improvements than some people do with their HDTV products," he claimed.

Xie said the software can be used with any codec and is useful for both consumer and professional video applications.

The startup conducted the first demos of the software a few weeks ago and is now ready to ship the code. Ipera charges a flat licensing fee, non-recurring engineering fees if users need customization of the code and a royalty based on the type and volume sales of the product using it.

The code has been optimized for an ARM 926EJS for use enhancing video playback in devices such as cellphones and digital cameras. It can provide enhancements to quarter VGA resolution video on handsets running on a 150 MHz or slower processor. It can also run on an ARM7-class CPU for lower-end feature phones.

The software is also optimized to run on an X86 processor for use in streaming video applications.

Pixel Magic can run under either Linux or Windows CE. The algorithm itself takes up just 100 Kbytes of storage.

The company's name refers to what Xie sees as an intellectual property (IP) era in which it licenses its algorithms to others who will provide the base hardware to run it. "At this stage we are trying to refrain from developing an ASIC ourselves, but we haven't made a final decision whether or not we will go into the silicon business," Xie said.

Ipera has raised less than three million dollars since it was launched in September 2005. It would require more than ten million more if it decides to develop its own chip, Xie said.

The startup does plan before the end of the year to release a reference design based on its code running in an FPGA. "It would be best to implement this software in a media processor," Xie said.