PTAB

IPR2024-01141

Amazon.com Inc v. Nokia Technologies Oy

Log In
Key Events
Petition
petition

1. Case Identification

2. Patent Overview

  • Title: Sub-pixel Value Interpolation
  • Brief Description: The ’273 patent discloses methods for sub-pixel value interpolation used in video compression. The technology determines values for virtual sub-pixels situated between actual pixels to improve the accuracy of motion compensation in video encoding and decoding.

3. Grounds for Unpatentability

Ground 1: Obviousness over TML6 and Fandrianto - Claims 1, 4-17, 19-26, 28-29, 31, 33, 36-49, 51, 53-59, 61-62, 64-78, and 82-83 are obvious over TML6 in view of Fandrianto.

  • Prior Art Relied Upon: TML6 (H.26L Test Model Long Term Number 6) and Fandrianto (Patent 5,594,813).
  • Core Argument for this Ground:
    • Prior Art Mapping: Petitioner argued that TML6, a publicly available source code for a video codec, discloses nearly all elements of the challenged claims, including a method for sub-pixel value interpolation for motion compensation. TML6 teaches interpolating quarter-quarter sub-pixels using a weighted average of four values: the nearest unit pixel and three nearest half-sub-pixels. Petitioner contended that Fandrianto, which addresses rapid and efficient motion vector searching, teaches an alternative interpolation method suitable for limited-memory hardware. Specifically, Fandrianto teaches interpolating a quarter-pixel value using a weighted average of only two values, such as a pair of diagonally-opposed sub-pixels. Petitioner asserted that the combination of TML6 and Fandrianto teaches the claimed invention, particularly the limitation in claim 1 requiring a "predetermined choice" of interpolation methods, including one based on a pair of diagonally-opposed sub-pixels.
    • Motivation to Combine: Petitioner argued a person of ordinary skill in the art (POSITA) would combine TML6 and Fandrianto because they are in the same technical field (video coding), address the same problem (sub-pixel motion estimation), and provide complementary solutions. Fandrianto's method of using two values for interpolation is less computationally intensive than TML6's method using four values. A POSITA would have been motivated to substitute Fandrianto’s computationally efficient interpolation technique into the TML6 framework as a simple design choice to improve performance, particularly in hardware with limited memory, a consideration explicitly mentioned by Fandrianto.
    • Expectation of Success: Petitioner asserted a POSITA would have had a reasonable expectation of success. The combination involves substituting one known interpolation method for another using the same inputs (pixel and sub-pixel values) and basic arithmetic. TML6 already calculates and stores the necessary input values (unit and half sub-pixels) required by Fandrianto’s method. Therefore, implementing Fandrianto’s technique in the TML6 system would have been a straightforward modification yielding the predictable result of a less computationally complex interpolation.

Ground 2: Obviousness over TML6, Fandrianto, and POSITA Knowledge - Claims 32, 52, and 79-81 are obvious over TML6 in view of Fandrianto and the knowledge of a POSITA.

  • Prior Art Relied Upon: TML6, Fandrianto (’813 patent), and the general knowledge of a POSITA.
  • Core Argument for this Ground:
    • Prior Art Mapping: This ground extended the arguments from Ground 1 to claims directed to "eighth resolution" (1/8th pixel) interpolation, corresponding to N=3 in the claims' coordinate notation. Petitioner argued that while TML6 discloses quarter-resolution (N=2), modifying the combined TML6/Fandrianto system to achieve eighth-resolution was a simple and obvious step for a POSITA.
    • Motivation to Combine (and Modify): Petitioner contended that by 2001, increasing interpolation accuracy to eighth-pixel resolution was a well-known technique for enhancing digital image quality and was a known trade-off between improved accuracy and computational complexity. The draft successor to TML6, known as TML8, had already implemented 1/8 pixel accuracy for higher efficiency profiles. A POSITA would have been motivated to apply this known technique of higher-order interpolation to the known TML6/Fandrianto device to achieve the predictable result of improved video fidelity.
    • Expectation of Success: Petitioner argued that setting N=3 instead of N=2 in the TML6/Fandrianto system would have been a trivial modification for a POSITA. The underlying interpolation principles remain the same, and the modification would result in a system that worked in the same way but with higher resolution. Eighth-resolution interpolation was achievable with hardware available at the time, making it a practical and predictable improvement.

4. Arguments Regarding Discretionary Denial

  • Petitioner argued against discretionary denial under General Plastic, Fintiv, and §325(d).
    • General Plastic: Petitioner asserted this is its first IPR petition against the ’273 patent and that it lacks the "significant relationship" with a prior petitioner (Lenovo) that would warrant denial. The separate infringement suits against Amazon and Lenovo involved different products and were filed years apart.
    • Fintiv: Petitioner argued that the co-pending district court case was in its earliest stages, with no trial date set, making a final resolution likely to occur well after the Final Written Decision (FWD) deadline for this IPR.
    • §325(d): Petitioner contended that the Examiner did not substantively consider the key prior art during prosecution. While the ’273 patent specification briefly summarizes TML6, it lacks the technical detail of the full TML6 source code relied upon in the petition. Furthermore, Fandrianto was only cited in an IDS and was never applied or discussed by the Examiner, and the specific combination of TML6 and Fandrianto was never considered.

5. Relief Requested

  • Petitioner requests institution of an inter partes review and cancellation of claims 1, 4-17, 19-26, 28-29, 31-33, 36-49, 51-59, 61-62, and 64-83 of the ’273 patent as unpatentable.