Monday, April 26, 2010

CVision Introduces SuperFast OCR

CVision, a Queens, NY-based document capture software developer, with some serious roots in JBIG 2 and compression technology, showed its new SuperFast OCR technology at last week's AIIM show in Philadelphia. The feature will initially be included in CVision's PDFCompressor flagship application for creating optimally compressed PDF files from images. Traditionally, CVision has licensed Nuance's OCR technology to make the PDFs full-text searchable. CVision will still utilize Nuance's OCR, but its the pre-processing which makes CVision's OCR so special.

From what I understand, CVision is able to convert recognized characters into symbols, which greatly accelerates the OCR process, especially on multi-page documents. The demo I saw processed 400 pages in 18 seconds, or an average of 22 pages per second. CEO Ari Gross told me CVision has tested the technology to deliver 10 pages per second per core processor. "With a quad-core processors, we can hit 40 pages per second," he said. Average OCR rates, from what I understand, are at least several seconds per page.

So, where does SuperFast OCR make the most sense? According to Gross, it's as an embedded feature on MFPs and scanners. Gross says it creates the potential for "real-time" OCR on these devices. In other words, as fast as the device outputs pages, it would be able to OCR them. Gross is gauging interest on embedding CVision's technology.

We recently wrote a story about Konica Minolta embedding ABBYY's OCR technology in its MFPs. However, Dean Tang, president of ABBYY USA, indicated that for users capturing and apply OCR to multi-page documents, it would probably be more efficient to apply the technology offline in a post-scan process. CVision's technology, which is in beta and will be released as part of PDFCompressor in a few weeks, could potentially eliminate the need to run a separate OCR process.


Chris Riley said...

Scanning at rated speed is such a cool development with the technology. Before the gap between inline OCR, that on firmware chips and PC based OCR in terms of performance was substantial. Although PC OCR has always been far more accurate and powerful, the performance for large volume scanning could be a hindrance. Getting closer to rated speed really helps. One aspect of CVisions demo that was most impressive to me however was not just the OCR, but really the compression.

Chris Riley

DIReditor said...

Thanks, Chris. CVision has always had great compression, which they licensed to Adobe. At the event, I saw a demo of the new imageRunner advance series, which looks like a pretty industrial MFP line, and it used some advanced compression - which I'm thinking was CVision's, but maybe not. Anyhow,I'm interested to see if they can have success licensing this stuff to hardware manufacture either through chip vendors or direct.

buyi wen said...

You can try this free online ocr tool, it can save the recognized text to searchable PDF file.