Quad Bayer Demosaicing Using Green-Channel Correlation

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing image processing algorithms consume significant CPU bandwidth and power due to software execution, and existing demosaicing methods do not effectively utilize high-resolution green pixel data to enhance non-green pixel data in raw images.

Innovation Solution

Generate full-resolution green pixel data by interpolating green pixels in multiple directions and combine them with high-pass filtered versions to generate full-resolution red and blue pixel data using correlation values, applying weights to interpolate and refine non-green pixel data.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If image processing algorithms are executed as software programs on a CPU, then processing flexibility is maintained, but CPU bandwidth consumption and power consumption increase significantly

Engineering Contradiction:

Improveprocessing flexibilityVSAvoidCPU power consumption

Core Design Contradiction:

Adaptability or versatilityVSUse of energy by moving object

Solution Approach 1:

The patent replaces software-based CPU processing with a dedicated hardware image processing pipeline that includes specialized circuitry for demosaicing operations. This hardware implementation maintains processing flexibility through configurable circuits while dramatically reducing CPU bandwidth consumption and power usage by offloading image processing tasks from the general-purpose CPU to purpose-built hardware components.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

2Measurement precision

If conventional demosaicing methods are used, then processing speed is maintained, but accuracy and resolution of non-green pixel data are insufficient

Engineering Contradiction:

Improvepixel data accuracyVSAvoidprocessing efficiency

Core Design Contradiction:

Measurement precisionVSProductivity

Solution Approach 1:

The patent performs preliminary generation of full-resolution green image data through multi-directional interpolation before using it to guide the reconstruction of non-green pixel data. This preliminary action creates a high-accuracy reference framework that enables subsequent non-green pixel reconstruction to achieve superior accuracy and resolution while maintaining processing efficiency through the hardware pipeline.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent introduces full-resolution green image data as an intermediary element that mediates between the raw Bayer pattern data and the final full-color output. By using the green channel data (which has full pixel coverage) as a reference and intermediary, the system can accurately reconstruct red and blue channel data through correlation analysis, thereby improving overall pixel data accuracy without sacrificing processing speed.

Inventive Principle:

Principle #24Intermediary (Mediator)

Data Source

PatentUS12563308B2Demosaicing quad bayer raw image using correlation of color channels

Publication Date: 2026.02.24 APPLE INC

AI summary

Embodiments relate to demosaicing a raw image into a full-resolution green pixels and using high-frequency components of the full-resolution green pixels to add details into red image pixels and blue image pixels. The full-resolution green pixels are generated by interpolating green pixels in the raw image using residual weights based on correlation between green pixels and non-green pixels along different directions, and then combining the green pixel values interpolated along different directions using weights derived from gradient information. The full-resolution red and blue pixels are generated by determining weights according to the correlation of interpolated red and blue pixels relative to the full-resolution green pixels, and adding the high-frequency components of the full-resolution green pixels to the interpolated red and blue pixels.