Obviousness

Obviousness Analysis of US Patent 11805267 under 35 U.S.C. § 103

This analysis identifies combinations of prior art that would render the independent claims of US Patent 11805267 obvious to a person having ordinary skill in the art (POSITA) prior to the patent's effective filing date (priority date of January 7, 2011).

The Inventive Concept

The core inventive concept of US11805267, as articulated in its independent claims (Claims 1, 13, 25, 26, 27, and 35), is to reduce rounding errors in bi-directional or multi-directional motion prediction within video coding. This is achieved by maintaining the prediction signals (first and second predictions) at a "second precision" (a higher precision) during intermediate calculations and only reducing their precision to a "first precision" (the final output precision) after these higher-precision prediction signals have been combined. A stated benefit of this approach is the removal of the need to signal rounding direction in the bitstream.

Identified Prior Art References

1. H.264/AVC Video Coding Standard (ITU-T Rec. H.264, ISO/IEC 14496-10): This widely adopted video coding standard, published in 2003, predates the patent's priority date and is explicitly mentioned in the background of US11805267 as conventional technology. H.264/AVC teaches fundamental concepts in video coding, including:

   • Block-based coding and prediction types: It defines intra-coded (I), uni-predicted (P), and bi-predicted (B) pictures and blocks, and methods for determining block types.
   • Motion compensated prediction (MCP): It extensively describes how a prediction for a current block is formed using previously coded frames.
   • Fractional-pixel motion vector accuracy: H.264/AVC supports motion vectors with up to quarter-pixel accuracy, requiring interpolation filters (e.g., symmetric and separable 6-tap filters for half-pixel samples, and averaging for quarter-pixel samples).
   • Bi-directional prediction: The standard specifically utilizes bi-directional motion compensated prediction, where a prediction signal for a block may be formed by combining (e.g., averaging) two motion compensated prediction blocks. The patent's background describes how such averaging operations in conventional codecs (like H.264/AVC) may include "up or down rounding, which may introduce rounding errors." It also mentions that "rounding error accumulation may be removed or decreased by signalling whether rounding up or rounding down have been used when the two prediction signals have been combined for each frame."

2. General Principles of Digital Signal Processing (DSP): Prior to 2011, it was well-established and generally known in the field of digital signal processing that maintaining a higher precision (i.e., more bits) for intermediate calculations in a processing chain helps to mitigate the accumulation of quantization, truncation, or rounding errors. This higher precision is then reduced to the desired output precision only at the final stage of calculation. This principle is fundamental to preserving accuracy in numerical computations, particularly in filtering, summation, and other arithmetic operations performed on sampled data.

Obviousness Argument

A POSITA in the field of video coding, prior to January 7, 2011, would have been acutely aware of the techniques described in the H.264/AVC standard. They would have been familiar with its use of bi-directional prediction, fractional-pixel interpolation, and the problem of rounding errors that could degrade coding efficiency, especially when combining prediction signals. The patent itself explicitly highlights this problem: "The accumulation of rounding errors in bi-directional prediction may cause degradation in coding efficiency." Furthermore, the existing solutions described in the patent's background, such as signaling rounding direction, were known to "increase somewhat the complexity as two separate code branches need to be written for bi-directional averaging."

Given these known problems and complexities, a POSITA would have been motivated to find improved methods for reducing rounding errors in multi-directional prediction while also simplifying the coding process (e.g., by avoiding the need to signal rounding direction).

The combination that renders the claims of US11805267 obvious would be:

• H.264/AVC (Reference 1) teaching: The overall framework of block-based video coding, bi-directional prediction, determination of block types, identification of reference pixel locations using motion vectors, and the process of obtaining first and second predictions via interpolation filters (e.g., 6-tap filters). H.264/AVC also teaches the conventional practice of combining these predictions and that rounding errors accumulate when precision is reduced too early.
• General DSP principles (Reference 2) teaching: The widely known engineering principle of performing intermediate numerical calculations at a higher precision to minimize the accumulation of errors before a final downscaling to the target precision.

Motivation for Combination:

A POSITA, recognizing the problem of accumulated rounding errors in H.264/AVC's bi-directional prediction (as described in the background of US11805267) and the complexity introduced by managing explicit rounding directions, would have been motivated to apply the well-known DSP principle of maintaining higher intermediate precision. It would have been obvious to a POSITA to adapt the H.264/AVC bi-prediction process by:

1. Performing the interpolation and initial prediction signal generation (e.g., P1 and P2 in FIG. 11) using a higher bit-depth (second precision) to ensure greater accuracy, consistent with general DSP best practices for multi-stage calculations.
2. Delaying the final precision reduction (to the first precision) until after the individual, higher-precision prediction signals have been combined. This directly addresses the problem of error accumulation that occurs when individual prediction signals are rounded prematurely (as depicted in FIG. 9's conventional approach).

By applying this known DSP technique to the H.264/AVC bi-prediction scheme, the POSITA would anticipate achieving the benefits of reduced rounding errors and, consequently, improved coding efficiency, as well as the simplification of the coding process by eliminating the need to signal rounding direction. The invention's solution is thus a straightforward application of a known principle to a known problem in a known system, yielding predictable results.