PTAB

IPR2025-01375

Snap Inc v. Nokia Technologies Oy

Log In
Key Events
Petition
petition

1. Case Identification

2. Patent Overview

  • Title: Utilizing motion prediction in video coding.
  • Brief Description: The ’267 patent describes methods for reducing rounding errors during bi-directional motion prediction in video coding. The disclosed technique involves maintaining prediction signals at a higher precision during intermediate calculations and then decreasing the precision after the signals have been combined.

3. Grounds for Unpatentability

Ground 1 & 1A: Obviousness over Wada and Karczewicz-I - Claims 1-36 are obvious over Wada in view of Karczewicz-I. Claims 3, 9, 15, 21, 27, and 33 are further obvious in view of Srinivasan.

  • Prior Art Relied Upon: Wada (Application # 2010/0322303), Karczewicz-I (Application # 2011/0007799), and Srinivasan (a 2005 SPIE proceeding).
  • Core Argument for this Ground:
    • Prior Art Mapping: Petitioner argued that Wada disclosed a prediction image generator that increases the precision of pixel values (e.g., from N bits to M or L bits) during intermediate processing stages before decreasing the precision back to N bits for the final output. Wada explicitly taught that its generator uses weighted prediction equations from the AVC/H.264 standard. Karczewicz-I was presented as teaching the specific default bi-directional prediction equations used in the AVC/H.264 standard, which involve averaging two predictions from different reference blocks. The combination of Wada's high-precision framework with Karczewicz-I's standard bi-prediction equations allegedly disclosed obtaining first and second predictions at a higher precision, combining them, and then right-shifting the result to decrease the precision. For Ground 1A, Srinivasan was added for its teachings on managing intermediate precision in sub-pixel interpolation by right-shifting the sum of a P-tap filter to ensure intermediate values fit within 16-bit registers.
    • Motivation to Combine: A POSITA would combine Wada and Karczewicz-I because Wada explicitly directed using prediction equations from the AVC standard, which Karczewicz-I described in detail. A POSITA implementing Wada’s system would naturally consult a reference like Karczewicz-I for the standard equations. Srinivasan would have been combined to address the well-known problem of managing bit-depth growth during multi-stage filtering, offering an established method to handle the higher-precision intermediate values generated in Wada's system during sub-pixel interpolation.
    • Expectation of Success: Petitioner asserted a POSITA would have an expectation of success because the combination involved applying standard H.264 equations (from Karczewicz-I) to a system designed to use them (Wada), representing a straightforward implementation rather than a novel invention.

Ground 2 & 2A: Obviousness over Karczewicz-I and Karczewicz-II - Claims 1-36 are obvious over Karczewicz-I in view of Karczewicz-II. Claims 3, 9, 15, 21, 27, and 33 are further obvious in view of Srinivasan.

  • Prior Art Relied Upon: Karczewicz-I (Application # 2011/0007799), Karczewicz-II (Application # 2009/0257499), and Srinivasan (a 2005 SPIE proceeding).
  • Core Argument for this Ground:
    • Prior Art Mapping: Petitioner contended that Karczewicz-I taught conventional H.264 bi-directional prediction, which averages two predictions (e.g., from interpolated subpixels) and applies a simple rounding adjustment. This method was susceptible to "propagation of rounding inaccuracies" if the input values were rounded prematurely. Karczewicz-II addressed this very problem in the similar context of quarter-pixel interpolation by teaching the use of higher-precision, non-rounded intermediate half-pixel values. Karczewicz-II disclosed specific equations for averaging pixel values of different precisions by left-shifting lower-precision values to match, using a larger rounding offset, and applying a larger right-shift to return the final value to the standard 8-bit precision. Petitioner argued it would be obvious to apply the precision-improving techniques of Karczewicz-II to the analogous bi-prediction averaging process in Karczewicz-I. Srinivasan (Ground 2A) was added for the same reasons as in Ground 1A.
    • Motivation to Combine: A POSITA would combine these references because they were from the same inventors, addressed nearly identical H.264 coding architectures, and solved a common problem (rounding errors). Petitioner highlighted the mathematical similarity between the quarter-pixel interpolation formula improved by Karczewicz-II and the bi-prediction formula from Karczewicz-I. A POSITA would have recognized that the technique used to improve one could be directly applied to improve the other.
    • Expectation of Success: The combination was asserted to be predictable because it involved applying the specific, well-defined mathematical operations from Karczewicz-II to the corresponding calculations in Karczewicz-I. Given the similar technical context and shared architectural components, a POSITA would expect the combination to work as intended to produce more accurate prediction values.

4. Key Claim Construction Positions

  • "precision": Petitioner argued that, in the context of the ’267 patent, "precision" should be construed to mean "the number of bits needed to represent possible values." This construction was presented as consistent with the specification, which describes right-shifting to make precision become "M bits." This position was framed to counter a potential, narrower construction by the Patent Owner in related litigation that might exclude certain prior art teachings.

5. Relief Requested

  • Petitioner requests institution of an inter partes review and cancellation of claims 1-36 of the ’267 patent as unpatentable.