Obviousness

US Patent 10218995 focuses on improving moving picture encoding efficiency, particularly in hierarchical encoding systems, by incorporating super-resolution enlargement and resolution conversion techniques. The patent explicitly identifies aspects of "Patent Literature 1 and MPEG-4 SVC" as relevant prior art that utilize inter-layer prediction to enhance encoding efficiencies by exploiting correlations between different spatial resolutions. However, the patent notes a problem with these prior art techniques: "failures to allot a sufficient code rate would degrade image qualities of predictive pictures or predictive blocks, resulting in reduced encoding efficiencies."

For an obviousness analysis under 35 U.S.C. § 103, one must consider whether the claimed invention as a whole would have been obvious to a person having ordinary skill in the art (PHOSITA) at the time of the invention, given the prior art. This includes identifying a motivation to combine prior art references.

Identified Prior Art References:

1. Patent Literature 1 and MPEG-4 SVC: These are described as prior art that "include predictive pictures or predictive blocks created from reference pictures of a layer lower than the layer in which a current encoding is made, for use in combination with target pictures or target blocks of the current layer to make an inter-layer prediction in between, aiming at still enhanced encoding efficiencies making use of high correlations between different spatial resolutions."
2. MPEG-2 and JPEG2000: These are mentioned in the patent as examples of typical motion picture encoding or pre-super-resolution encoder elements, but without specific disclosures relevant to forming a detailed obviousness combination in this context.

Core Inventive Concepts of US10218995:

The patent describes a moving picture encoding system that includes, among other things:

• A first encoder that encodes and decodes moving pictures with a standard resolution.
• A first super-resolution enlarger that works on the standard resolution input pictures to create super-resolution enlarged pictures with a higher resolution.
• A first resolution converter that converts these super-resolution enlarged pictures, often back to the standard resolution, to create "super-resolution enlarged and converted pictures."
• A second encoder that uses these "super-resolution enlarged and converted pictures" as encoding target pictures. Crucially, the second encoder may also use "decoded pictures" from the first encoder as a first set of reference pictures and/or "super-resolution enlarged and converted decoded pictures" (derived by processing the first encoder's decoded output through a second super-resolution enlarger and a second resolution converter) as a second set of reference pictures for prediction.

The objective is to allow the second encoder to encode moving pictures "based on an increased amount of information relative to an information amount of moving pictures input with the standard resolution".

Obviousness Combinations and Motivation:

A plausible combination of prior art that could render certain claims of US10218995 obvious to a PHOSITA would involve combining the principles of hierarchical inter-layer prediction (as taught by MPEG-4 SVC) with known super-resolution enlargement and resolution conversion techniques.

Hypothetical Combination:

1. Primary Reference: MPEG-4 SVC (or Patent Literature 1)

   • Disclosure: This reference teaches hierarchical video encoding where predictive pictures or blocks from a lower resolution layer are used to predict target pictures or blocks in a higher (or current) resolution layer. This leverages spatial correlations to improve encoding efficiency.
   • Problem Addressed: The patent itself identifies that such techniques are prone to degraded image quality and reduced encoding efficiencies when sufficient code rates are not allotted.

2. Secondary Reference: Known Super-Resolution and Resolution Conversion Techniques

   • Disclosure: At the time of the invention's priority date (May 30, 2008), super-resolution enlargement was a known technique for reconstructing higher-resolution images from one or more lower-resolution images, often by leveraging sub-pixel shifts or temporal information. The patent itself describes a super-resolution enlarger (e.g., 103) with components like a positioner, interpolator, and estimated picture creator, which are typical elements of super-resolution processes. Similarly, resolution conversion (upsampling/downsampling with filtering) was a well-established image processing technique, as evidenced by the components of the first resolution converter 104 (pixel inserter, filtering processor, pixel thinner).

Motivation for Combination:

A PHOSITA, faced with the recognized problem in MPEG-4 SVC and Patent Literature 1—namely, the degradation of image quality and reduced encoding efficiency in inter-layer prediction when code rates are limited—would have been motivated to seek solutions to provide more robust or informative reference signals for prediction.

• To Improve Prediction Quality and Efficiency: The explicit motivation for super-resolution techniques is to recover or infer higher frequency components and detail that may be latent in lower-resolution input, effectively increasing the "amount of information." Given that MPEG-4 SVC's inter-layer prediction relies on correlations between spatial resolutions and aims for enhanced efficiency, a PHOSITA would naturally consider methods to enrich the information content of pictures used in this prediction.
• Leveraging Latent Information: Super-resolution enlargement is designed to "includ[e] information on frequency components in the spatial direction and the temporal direction that has been potentially contained in the input moving pictures but unable to express to a sufficient degree by the standard resolution". Applying such a technique to the input pictures, or to decoded pictures, prior to or within the hierarchical encoding process, would be a logical step to provide the inter-layer prediction mechanism (e.g., in the second encoder) with a "greater amount of information" to work with, thereby directly addressing the efficiency and quality issues identified by the patent itself.
• Resolution Matching and Bandwidth Control: The subsequent resolution conversion step (e.g., bringing the super-resolution enlarged picture back to the standard resolution) would be motivated by a need to ensure compatibility with the standard resolution layer of the hierarchical encoding system while still retaining the benefits of the enhanced frequency information. The patent notes the resolution converter's function to "restrict within a range of information on frequency components that can be expressed with the standard resolution, to include in the super-resolution picture". This ensures the benefits of super-resolution are reaped within the constraints of the target resolution.
• Adaptive Prediction: The idea of providing multiple reference pictures (e.g., a simply decoded picture and a super-resolution processed decoded picture) to the second encoder, allowing for selection based on optimality, would be a straightforward engineering decision for a PHOSITA seeking to maximize prediction accuracy and encoding efficiency.

Therefore, a PHOSITA, understanding the limitations of existing inter-layer prediction in scalable video coding like MPEG-4 SVC concerning code rate efficiency and image quality, and aware of the capabilities of super-resolution and resolution conversion to enhance image information, would have had a clear motivation to combine these techniques to provide richer and more effective reference or target pictures for the prediction process, leading to the improvements claimed by US10218995. This combination would be a predictable solution to a known problem in the art.