Google LLC v. Uniloc 2017 LLC

1. Case Identification

• Case #: IPR2020-00447
• Patent #: 6,329,934
• Filed: January 17, 2020
• Petitioner(s): Google LLC
• Patent Owner(s): Uniloc 2017 LLC
• Challenged Claims: 1

2. Patent Overview

• Title: Method of Modifying Data
• Brief Description: The ’934 patent discloses a method and apparatus for modifying coded data, such as an MPEG video stream. The purported invention involves a three-stage process of partially decoding a coded data stream, modifying the resulting partially decoded data, and then performing a complementary coding step to re-encode the modified data.

3. Grounds for Unpatentability

Ground 1: Claim 1 is anticipated under 35 U.S.C. §102(e) by Liu.

• Prior Art Relied Upon: Liu (Patent 5,907,374).
• Core Argument for this Ground:
  • Prior Art Mapping: Petitioner argued that Liu discloses every limitation of the challenged method claim. Liu describes a system for processing compressed bitstreams that includes a layered decoder, a decoded bitstream processor, and a layered encoder. Petitioner asserted that Liu’s layered decoder (114) performs the claimed "partially decoding" by stopping the decoding process after certain steps (e.g., after inverse discrete cosine transform) but before others (e.g., motion compensation) to produce prediction-error pixels. Liu’s decoded bitstream processor (118) then performs the claimed "modifying" step on this partially decoded data through operations like color correction, scaling, or filtering. Finally, Liu’s layered encoder (124) performs the claimed "complementary coding" by re-encoding the modified data using only the encoding steps that correspond to the decoding steps that were performed, thus generating a modified coded output stream.

Ground 2: Claim 1 is obvious over Liu in view of Chitprasert.

• Prior Art Relied Upon: Liu (Patent 5,907,374) and Chitprasert (“Discrete Cosine Transform Filtering,” Signal Processing, 1990).
• Core Argument for this Ground:
  • Prior Art Mapping: Petitioner presented this ground as an alternative, arguing that if Liu is found not to explicitly teach modifying the blocks of prediction-error pixels in the spatial domain, the combination with Chitprasert renders this step obvious. Liu provides the foundational three-stage system of partial decoding, modification, and complementary re-encoding. Chitprasert teaches that filtering image data can be performed equivalently in either the frequency domain (on DCT coefficients) or the spatial domain (on pixels), and that the choice between them is a known design trade-off. Liu’s bitstream processor already modifies data, and Chitprasert teaches a well-known, interchangeable method for performing such modification (filtering) directly on pixel data.
  • Motivation to Combine: A POSITA would combine Chitprasert’s teaching of spatial domain filtering with Liu’s system to enable more flexible and precise image processing. Specifically, filtering in the frequency domain is an approximation of filtering in the spatial domain; a POSITA would have been motivated to implement direct spatial domain filtering in Liu's system for situations where such an approximation is undesirable. This would have been a simple modification to incorporate a known technique into an existing system to improve its capabilities.
  • Expectation of Success: A POSITA would have a reasonable expectation of success because combining these known elements was a predictable application of established techniques. Liu’s system was designed for bitstream modification, and Chitprasert provided a known filtering method. Implementing Chitprasert’s spatial filtering in Liu’s processor would predictably result in a system that filters partially decoded pixel data without disrupting the overall partial decoding and complementary encoding architecture.

4. Key Claim Construction Positions

Petitioner argued for specific constructions of key terms in claim 1, contending they were critical to the invalidity analysis.

• "partially decoding": Petitioner proposed this term be construed as "variable length decoding, inverse quantizing, and inverse discrete cosine transforming, but not motion compensating." This construction is central to the argument, as it defines the partially decoded data as prediction-error pixels in the spatial domain, distinguishing it from both fully decoded data (which would include motion compensation) and frequency-domain data (which would not have undergone an inverse transform).
• "modifying": While applying the plain and ordinary meaning, Petitioner also analyzed this term under a construction from a related district court proceeding: "altered blocks of prediction-error pixels." Petitioner argued the prior art met the limitation under either interpretation.
• "complementary coding": Petitioner proposed this term be construed as performing the inverse operations of the "partially decoding" step, specifically "discrete cosine transforming, quantizing, and variable length coding." This construction links the re-encoding process directly back to the specific steps of the partial decoding, ensuring a symmetric process as described in the ’934 patent.

5. Relief Requested

• Petitioner requests institution of an inter partes review and cancellation of claim 1 of the ’934 patent as unpatentable.