Obviousness

US patent 10250877 covers methods and devices for coding and decoding image blocks, particularly focusing on handling reference images of different sizes in adaptive resolution video coding. The core inventive concept across its independent claims (Claims 1, 4, 7, and 8) is the joint application of motion compensation interpolation filters (MCIF) and resampling filters (SCF) into single horizontal (GFH) and vertical (GFv) filters (e.g., GFv(s) = MCIFv(SCFv(s)) and GFH(u) = MCIFH(SCFH(u))), thereby eliminating the need to store resampled versions of reference images in a decoded picture buffer (DPB). This addresses the problem of excessive memory consumption in prior art adaptive resolution systems.

An analysis of obviousness under 35 U.S.C. § 103 indicates that the claimed invention would have been obvious to a person having ordinary skill in the art (PHOSITA) by combining existing prior art references. The critical date for prior art assessment is the earliest priority date of January 13, 2012.

Prior Art References and their Teachings:

1. US20100226437A1 (Sony Corporation), published September 9, 2010: This patent application discloses "Reduced-resolution decoding of AVC bit streams for transcoding or display at lower resolution". This reference demonstrates the known challenge of handling different resolutions in video decoding and the necessity for resolution adjustment.
2. Davies et al., "AHG18 adaptive resolution coding (ARC)," published November 21, 2011: This non-patent literature explicitly discusses adaptive resolution coding in the context of advanced video coding, indicating that the general problem of varying image sizes was well-recognized in the field. Other Davies et al. papers, "JCTVC AHG report resolution adaptation" (Nov 21, 2011) and "Resolution switching for coding efficiency and error resilience" (Jul 15, 2011), further reinforce the awareness of resolution adaptation and switching.
3. Han et al., "Joint optimization of the motion estimation module and the Up/Down scaler in transcoders," published October 1, 2005: This article teaches the "Joint optimization of the motion estimation module and the Up/Down scaler". Although presented in the context of transcoders, the principle of combining or jointly processing motion estimation (which involves interpolation) and scaling (up/down sampling) operations was established. This directly suggests a technical approach to integrate these two functions.
4. Wittmann et al., "Separable adaptive interpolation filter for video coding," published October 12, 2008: This paper describes "Separable adaptive interpolation filter for video coding," confirming that the use of adaptive and separable (horizontal and vertical) interpolation filters was known in the art for video coding applications. The patent itself acknowledges that "The 2D interpolation filters used for the motion compensation and the 2D upsampling and subsampling filters are generally separable filters."

Obviousness Argument:

The independent claims of US10250877 address the problem of excessive memory usage in the DPB due to storing multiple resampled versions of reference images when adaptive resolution coding is employed. The patent explicitly articulates this problem in its "Prior Art" section, stating that "The adaptive resolution method thus requires more memory space as several versions (reconstructed image and upsampled and/or subsampled versions) of a same reference image are stored. Increasing the memory size is a problem notably in the hardware implementations for which the surface of the component is notably critical."

A PHOSITA would have been motivated to combine the teachings of the aforementioned prior art references to overcome this known memory limitation.

• Understanding the Problem (from 1, 2, and the patent's own "Prior Art" section): A PHOSITA would be aware of adaptive resolution video coding (US20100226437A1, Davies et al.) and the associated need to either resample reference images to match the current image's size or store multiple resampled versions in the DPB, leading to increased memory requirements. The patent itself states that "More generally, each reference image is stored in the DPB memory with N different sizes" in the prior art.
• Motivation for Joint Processing (from 3): Faced with the memory problem, a PHOSITA seeking to reduce DPB size would naturally look for ways to eliminate intermediate storage. Han et al.'s teaching of "joint optimization of the motion estimation module and the Up/Down scaler" would provide a strong motivation and conceptual basis for combining the resampling and motion compensation interpolation steps. The idea of performing these operations together, rather than sequentially with intermediate storage, would be an obvious engineering choice when memory reduction is a key objective.
• Implementing the Combination (from 4 and general knowledge): Knowing that both motion compensation interpolation and resampling filters are commonly implemented as separable (horizontal and vertical) filters (Wittmann et al.), a PHOSITA would readily understand that these two types of filters could be composed into a single, combined filter. The mathematical operation of filter composition (e.g., GFv(s)=MCIFv(SCFv(s))) is a standard technique in signal processing. This combined filter could then be applied directly to the original-resolution reference image to generate the motion-compensated prediction block, thereby obviating the need for storing resampled versions in the DPB, as claimed in US10250877.

Therefore, the combination of a reference teaching adaptive resolution and the associated memory burden (e.g., US20100226437A1 or Davies et al., coupled with the patent's own description of prior art), with a reference teaching joint processing of motion estimation and scaling (Han et al.), and the general knowledge of separable interpolation filters (Wittmann et al.), would render the claimed invention obvious. A PHOSITA would be motivated to make this combination to achieve the known desire of reducing memory footprint in adaptive resolution video coding systems.