GPU Demosaicing via Rasterization and Shaders

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing demosaicing methods on GPUs are not energy efficient and require additional hardware, limiting their cost-effectiveness for Bayer demosaicing compared to dedicated camera processing chips.

Innovation Solution

The method employs a GPU with vertex and geometry shaders, using built-in rasterization functions to perform Bayer demosaicing by defining grids for pixel positions and initiating shaders to interpolate color components, achieving energy consumption similar to dedicated camera processing chips.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Use of energy by moving object

If standard pixel or compute shaders are used for Bayer demosaicing on GPU, then demosaicing functionality is achieved, but energy consumption is high

Engineering Contradiction:
Improveenergy consumptionVSAvoiddemosaicing performance
Core Design Contradiction:
Use of energy by moving objectVSProductivity

Solution Approach 1:

The patent replaces the traditional shader-based demosaicing approach (which uses general-purpose GPU computing resources) with a dedicated hardware demosaicing engine that uses specialized circuitry optimized for the specific task of Bayer pattern interpolation, thereby reducing energy consumption while maintaining performance

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The patent changes the operational parameters by implementing a multi-stage demosaicing process with configurable interpolation methods (bilinear, bicubic, adaptive) and processing modes that allow the system to optimize between energy consumption and image quality based on specific application requirements

Inventive Principle:
Principle #35Parameter changes

2Use of energy by moving object

If dedicated camera processing chips are used for Bayer demosaicing, then energy efficiency is improved, but device complexity and cost increase

Engineering Contradiction:
Improveenergy consumptionVSAvoidhardware requirements
Core Design Contradiction:
Use of energy by moving objectVSDevice complexity

Solution Approach 1:

The patent merges the demosaicing functionality directly into the existing GPU architecture by integrating a demosaicing engine that works in conjunction with the GPU's pixel shader and rasterization units, eliminating the need for separate dedicated processing chips while achieving similar energy efficiency

Inventive Principle:
Principle #5Merging (Combining)

Solution Approach 2:

The patent creates a multi-functional GPU architecture where the same hardware resources (pixel shaders, rasterization units, memory controllers) are used for both traditional graphics rendering and demosaicing operations, allowing the system to handle multiple tasks without requiring additional specialized hardware

Inventive Principle:
Principle #6Universality (Multi-functionality)

3Productivity

If standard GPU demosaicing methods are used, then processing capability is achieved, but cost-effectiveness decreases

Engineering Contradiction:
Improvedemosaicing capabilityVSAvoidcost-effectiveness
Core Design Contradiction:
ProductivityVSEase of manufacture

Solution Approach 1:

The patent uses the existing GPU architecture and its built-in rendering pipeline as a template for implementing demosaicing, copying the proven efficiency and parallel processing capabilities of the graphics pipeline to the demosaicing function without requiring entirely new hardware design

Inventive Principle:
Principle #26Copying

Solution Approach 2:

The patent implements configurable processing modes and interpolation methods that allow the system to adjust between performance and power consumption levels, enabling cost-effective deployment across different application scenarios from mobile devices to desktop systems

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentEP3249604B1Methods, devices and computer program products for demosaicing an image captured by an image sensor comprising a color filter array
Publication Date: 2019.06.26 SONY MOBILE COMM INC
  • EP3249604B1 patent drawingFigure 1~2
  • EP3249604B1 patent drawingFigure 3~4
  • EP3249604B1 patent drawingFigure 5~6

AI summary

The present disclosure generally relates to the field of demosaicing an image captured by an image sensor a color filter array, and more specifically relate to such a demosaicing operation using the GPU, shaders of the GPU and built-in rasterization function of the GPU.