PTAB reverses when Examiner interprets "each pixel of a color image" as referring to pixels in different images

Takeaway: The Applicant appealed the obviousness rejection of claims to digital image processing. The claim language at issue was: "for each of multiple given pixels of a color image, determining ..." The Examiner read this on pixels in different images, and the Applicant argued it was limited to processing pixels in the same image. The Examiner relied on Baldwin Graphic Systems, Inc. v. Siebert, Inc., The Applicant considered the context of the determining step as a whole, including the term "local neighborhood." The Board agreed with the Applicant and reversed the rejection. (Ex parte Staelin, PTAB 2013)

Details:
Ex parte Staelin
Appeal 2012000632; Appl. No. 11/280,097; Tech. Center 2600
Decided:  February 28, 2013

The application on appeal was directed to digital image processing. A representative claim on appeal read:

     A method, comprising:
     for each of multiple given pixels of a color image having multiple color channels, determining a respective local difference measure value from a value of the given pixel and a respective value of at least one other pixel in a local neighborhood of the given pixel for each of multiple of the color channels;
     for each of the given pixels, ascertaining a respective color correlation value measuring deviation of the respective local difference measure values determined for the multiple color channels  from a reference: and
     calculating a level of noise in the color image from one or more of the ascertained color correlation values;
     wherein the determining, the ascertaining, and the calculating are performed by a machine.
(Emphasis added.)

The claim limitation at issue on appeal was the "respective local difference" step emphasized above.
This independent claim was rejected as obvious using a combination of two references.

The Examiner asserted that this step corresponded to the SAD (Sum of Absolute Differences) calculation in the primary reference. Applicant argued that the SAD (Sum of Absolute Differences) measure disclosed in the reference did not correspond to the claimed step because it used pixels in different frames. Thus, according to the Applicant, "the SAD measure does not constitute 'a respective value of at least one other pixel in a local neighborhood of the given pixel' " as claimed.

In a later Office Action, the Examiner gave this additional explanation:

However, the Examiner notes that "determining a respective local difference measure value [...] " is taught by Hamosfakidis by virtue of the fact that Hamosfakidis finding difference between pixels in a local neighborhood be it for a single image or of different frames as the Applicant argues.

The Applicant appealed. In the Appeal Brief, the Applicant picked up on the Examiner's explanation and asserted that as properly construed, the "local difference" is determined based on pixels in the same color image. The Applicant made a detailed claim construction argument, using dictionary definitions and focusing on the plain meaning of "local" and of "neighborhood." The Applicant concluded that "the plain meaning of the 'determining' element of claim 1 does not support the Examiner's position that the measurement of differences between  pixels of different frames."

The Examiner responded in the Answer as follows:

The examiner notes that the applicant does not claim that the determining step is done for the "same color image"... [T]he applicant does claim "a color image" but the examiner notes that in for example  in KJC Corp. v. Kinetic Concepts, Inc., 223 F.3d 1351, 1356 (Fed. Cir. 2000) and in Baldwin Graphic Systems, Inc. v. Siebert, Inc., the court confirmed that the use of the indefinite articles "a" or "an", in an open-ended claim containing the transitional phrase "comprising", carries the meaning of "one or more."

The Board found the Examiner's claim construction to be unreasonable:

We agree with Appellants that the Examiner's construction is unreasonable. The language of claim 1 recites ''for each of multiple given pixels of a color image having multiple color images." Therefore, we interpret the claim language of the "determining" step to require that the local distance measure be determined in the same color image. As noted by Appellants, Hamosfakidis teaches measuring differences between different frames,  i.e. different color images. Therefore, Hamosfakidis cannot teach taking a local difference value between a given pixel and a one other pixel  in a local neighborhood  of the given pixel, in a color image.
(Emphasis in original.)

The Board then reversed the obviousness rejection based on the Examiner's unreasonable construction.

My two cents: This decision makes sense to me, but I wish the Board had given a better explanation. The Board's phrasing suggests that it's the presence of "each of" that leads to the conclusion that "image" is limited to a single image – despite the presence of "a." Which makes sense to me, although perhaps my thinking is colored by my background as a programmer.

The Applicant emphasized the terms "local" and "neighborhood" in concluding that the repeated pixel processing ("for each") was performed on the same image. The Board didn't comment on how persuasive this piece of the argument was.

Though I think the Board got this one right, I wouldn't be surprised to see other panels apply the "A means one or more" rule and affirm the rejection. Amending to "a single color image" would have forced the Examiner to change his rejection.