Patent dispute involves industry leaders

Leading companies in the machine-vision and image-processing industry should be engaged in driving market growth instead of legal entanglement.

In a recent press release, Cognex Corp. (Natick, MA) claims that the Matrox Imaging (Dorval, QC, Canada) Geometric Model Finder software, and possibly other components, shipped as part of the company's Matrox Imaging Library, infringes the claims of at least one Cognex patent. The company is seeking court action to stop Matrox from making and selling Geometric Model Finder or any other product found to infringe Cognex's patents, as well as to award unspecified damages.

In the press release, Bill Silver, Cognex's chief technology officer, is quoted as saying that he had "personally investigated the Matrox Geometric Model Finder and believes that it infringes claims held by Cognex." If Cognex should win, the case may set a precedent for cases against other companies with similar software products, such as those from Adept Technology (Livermore, CA) and Coreco Imaging (Billerica, MA).

The patent-infringement claim is yet another legal entanglement for Cognex. For years, the company has been rightfully battling the Lemelson Foundation (Incline Village, NV), an organization designed to "stimulate the US economy and secure its position in the global marketplace" by claiming patent rights on everything from medical scanners to machine-vision systems (see Vision Systems Design, April 2002, p. 5).

Cognex's argument with Matrox is not so clear-cut. As with most mathematical functions, there is a limited number of ways to perform geometric pattern-matching. This could involve, for example, finding the edges of an object by thresholding the images, connecting adjacent edges using a morphological operator, and computing the angles between edges. Once a series of segments and angles are computed, the information can then be tabulated. When an image rotates, its geometric features do not change, and, so, individual objects with different geometric features can be rapidly identified.

Can such "algorithms" be patented? It's like asking whether a bicubic polynomial warp can be patented or, for that matter, any mathematical formula. Here, there are two underlying schools of thought. There are those who believe that once developed or discovered, mathematical algorithms should be immediately placed in the public domain. They were centuries ago, such as when mathematician Baron Jean Baptiste Joseph Fourier developed the Fourier transform. People who believe in the public domain would argue that, had Baron Fourier patented his algorithm, it would have appeared for years as "This formulation is patent-protected by Baron Fourier." Free thinkers would find this scenario highly objectionable.

On the other hand, other innovators believe that patents protect them and their inventions from unscrupulous competition. They argue that after spending large amounts of R&D dollars, they should be protected by patents in the published disclosure of the methods and means of their inventions. After a set number of years, these inventions fall into the public domain, and others can freely copy the invention.

Cognex and Matrox are both well-respected companies in the machine-vision and image-processing industry. It is therefore highly unlikely that Matrox engineers reverse-engineered a Cognex product to build their Geometric Model Finder software. Because of this, one wonders how the patent infringement of complex mathematical algorithms can be effectively determined.

Nonetheless, once a patent has been granted, the invention is established. It cannot be invented again. In an economic downturn, however, one asks whether handing money to lawyers is in the best interest of both companies and the industry. Two of the leading companies in the machine-vision and image-processing industry should be engaged in driving market growth instead of legal entanglement. The court will eventually decide the outcome.

Andy Wilson
EDITOR
andyw@pennwell.com

More in Boards & Software