Obviousness

Obviousness Analysis under 35 U.S.C. § 103 for US Patent 10,944,901

This analysis evaluates the obviousness of US Patent 10,944,901 ("the '901 patent") by identifying combinations of prior art references that would render the claimed invention obvious to a person having ordinary skill in the art (POSITA).

1. Closest Prior Art References

The '901 patent itself identifies several relevant prior art documents in its "BACKGROUND" section. For this analysis, we will focus on the following key references:

• US 2013/0155474 A1 (Sargent et al.): This reference discloses a system that provides feedback to a user prior to image capture, combines measured parameter values into an overall quality score, and captures an image only if this score exceeds a defined threshold. It also provides detailed instructions for the user to adjust parameters to achieve a better quality image, for instance, suggesting the user hold the camera steadier to address motion blur. [cite: a]
• WO 2006/040761 A1 / US 2007/0195174 A1 (Elbaz et al.): This reference describes a system where an interface module enables defining scene dynamics, including image motion speed, and allows setting relative weights for captured image attributes in the computation of a total image grade. [cite: b]
• US 2009/0278958 A1 (Choi et al.): This reference teaches that the scoring of a current image can be based on scores given to previously captured base images to avoid redundant calculations. [cite: c]

2. Differences Between the Claims of US 10,944,901 and the Closest Prior Art

The independent claims of the '901 patent (Claims 1, 16, and 17) introduce several features not explicitly present or combined in an obvious manner in the identified prior art:

• Confidence Level for Quality Indicators (QIs): The '901 patent explicitly claims computing "a confidence level for at least one of said quality indicators" (Claim 1) and utilizing these confidence levels in the overall system (Claim 16). The patent explains that prior art computed QIs "without taking into account the possibility of error in the computed QI, as all sensors give out errors," and that a "confidence level" is obtained to address unreliable or fluctuating QI values.
• Dynamic and Inter-Dependent Weighting of QIs: The '901 patent highlights that "the weight of one indicator will take into account data from other quality indicator/s e.g. their quality indicator value, weight, confidence level... and their previous value, weight and confidence level." This is contrasted with prior art that uses "constant or can be change by the user manually" weights. [cite: b] A key example provided is disregarding an aesthetic quality indicator if device shake or camera focus QIs indicate poor quality, even if the user prioritized aesthetics, to prevent blurry pictures.
• Automatic Adjustment of Lens/Sensor Module Parameters: Claim 1 specifies "adjusting parameters in the lens/sensor module to achieve better total QI." While Sargent et al. provides suggestions to the user for adjustment [cite: a], it does not teach automatic, system-initiated adjustment of camera hardware parameters (e.g., ISO, aperture, shutter speed, focus point) based on computed QIs and confidence levels.

3. Motivation to Combine Prior Art References

A POSITA, typically an engineer or scientist skilled in image processing, computational photography, or embedded systems for cameras, would aim to develop more robust, accurate, and user-friendly real-time photo quality assessment systems.

Combination of Sargent et al. and Choi et al.

• Motivation: To improve the reliability and temporal consistency of real-time image quality assessment. Sargent et al. provides a real-time system that assesses quality and provides feedback/takes action on a current frame. [cite: a] Choi et al. teaches using scores from previously captured images for current scoring to avoid redundant calculations. [cite: c] A POSITA would be motivated to combine these to enhance the quality assessment by considering not just the current frame but also historical data, thereby stabilizing the quality metric against transient fluctuations and producing a more reliable assessment over time. This would lead to a more effective real-time system that better assists users in capturing good images.

Adding Elbaz et al.

• Motivation: To allow for more flexible and context-aware determination of total image quality. Elbaz et al. teaches setting relative weights for different image attributes in computing a total image grade. [cite: b] A POSITA, having combined Sargent et al. and Choi et al. for a more stable quality assessment, would find it obvious to incorporate dynamic weighting (as taught by Elbaz et al.) to adapt the overall quality score to different scene conditions or user preferences, thereby making the system more versatile and intelligent.

Addressing Automatic Adjustment of Lens/Sensor Module Parameters

• Motivation: To enhance the automation and responsiveness of the image acquisition system. Sargent et al. provides detailed feedback to the user on how to improve an image (e.g., "hold the camera steadier"). [cite: a] Given the trend towards greater automation in modern camera systems (even at the priority date of 2012-10-23), a POSITA would be motivated to replace these user instructions with automatic adjustments of the lens/sensor module parameters (e.g., shutter speed, ISO, aperture, focus) to directly correct detected quality issues. This would improve the user experience by reducing manual intervention and enabling faster, more precise corrections in real time. For example, if the system detects excessive device shake (a parameter measured by Sargent et al. [cite: a]), it would be an obvious engineering solution to automatically increase ISO or change aperture to allow a faster shutter speed, rather than merely telling the user to "hold still." The patent also mentions using quality indications to control the lens module, for instance, shortening shutter speed if movement is detected perpendicular to the focus plane.

Addressing Confidence Levels and Inter-Dependent Dynamic Weighting

This is the most challenging aspect to render obvious. While general knowledge recognizes that sensor data and algorithmic outputs have errors, the explicit and detailed mechanism for:

1. Computing a specific "confidence level" (C_i(t)) for each individual quality indicator (e.g., using a probability factor P_i(t) and a normal distribution N(x) as shown in the patent's formula C_i(t) = P_i(t_j2) * N(P_i(t_j2))).
2. Using these confidence levels to dynamically adjust the overall QI value (QI_ForTotal(t)i = QI(t)i * Π f_ij(t, QI(t)j, Cj)).
3. Applying a weighting scheme (w(t, Ci)i) where the weight of one indicator explicitly "take[s] into account data from other quality indicator/s e.g. their quality indicator value, weight, confidence level... and their previous value, weight and confidence level." [cite: b]

None of the cited prior art references (Sargent et al., Elbaz et al., Choi et al.) explicitly teach this sophisticated, multi-factor, inter-dependent dynamic weighting scheme based on individually computed confidence levels for each QI. While Elbaz et al. mentions relative weights [cite: b], it does not suggest that the weight of an aesthetic QI should be disregarded if a blur QI indicates poor quality with high confidence, as specifically exemplified in the '901 patent. This conditional, confidence-aware weighting represents a significant step beyond merely assigning static or simple dynamic weights. The motivation to implement such a complex system would be to overcome the problem of prior art systems taking blurry pictures despite user preferences, by prioritizing fundamental quality issues over secondary ones when confidence in the fundamental issue is high. However, arriving at this specific solution without the explicit teaching would likely require more than routine experimentation or common sense from a POSITA.

4. Conclusion

The combination of Sargent et al., Choi et al., and Elbaz et al. would render obvious many aspects of the '901 patent's claims, including:

• Computing a total quality indicator from multiple QIs, including data from previous frames.
• Determining acceptability based on this total QI.
• Taking differential action (e.g., capturing the image or providing feedback).
• Providing detailed feedback to the user.
• Automatically adjusting camera parameters (e.g., ISO, shutter speed, aperture) in response to detected quality issues.

However, the specific methods for:

1. Explicitly computing a "confidence level" for each individual quality indicator, using detailed formulas involving probability factors and statistical distributions.
2. Dynamically adjusting the weight of one quality indicator based on the values and confidence levels of other quality indicators (and their previous states), particularly to conditionally disregard certain QIs (like aesthetic QIs when a fundamental flaw like shake or focus is present and highly confident).

These two features, as detailed in the '901 patent, represent a more sophisticated and intelligent approach to real-time quality assessment than what is explicitly taught or rendered obvious by the combination of the cited prior art. The patent itself highlights these as key differentiating aspects over prior art's limitations. Therefore, while many elements might be obvious, the specific, inter-dependent, and confidence-level-driven dynamic weighting scheme for quality indicators, particularly its explicit formulation and application, may present a stronger argument against obviousness.