Automatic Image Adjustment via Generic Label Translation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current image adjustment techniques are tedious and often fail to address complex adjustments, as they are subjective and cannot be directly embedded into algorithmic procedures, and different image processing tools use varying pipelines and color spaces, leading to inconsistent results.
Innovation Solution
The development of techniques to generate generic labels that are translated into pipeline-specific labels, allowing for automatic image adjustments by training a regression algorithm on image pairs and using these labels to adjust new images within various image processing pipelines.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If rule-based automatic techniques are used to adjust photographs, then the process becomes automated and faster, but the adjustment quality fails in complex cases such as back-lighting and difficult lighting situations
Solution Approach 1:
The patent introduces an intermediary machine learning model that acts as a bridge between automatic processing and manual adjustment quality. The model is trained on photographer adjustments to learn complex decision-making patterns, enabling automatic techniques to achieve high-quality results in difficult lighting situations while maintaining fast processing speeds.
Solution Approach 2:
The patent transforms the adjustment process by changing from fixed rule-based parameters to dynamic parameters learned from training data. The machine learning model adjusts multiple parameters (exposure, contrast, highlights, shadows) based on learned patterns from photographer corrections, enabling adaptive quality improvement across different lighting conditions.
2Adaptability or versatility
If different image processing pipelines with varying color spaces are used, then tool compatibility increases, but consistent mapping and translation between pipelines becomes difficult
Solution Approach 1:
The patent creates a universal adjustment representation that can be applied across different image processing pipelines. By training the machine learning model to output adjustments in a pipeline-agnostic format, the system achieves consistent translation between different color spaces (RGB, LAB, CMYK) and processing workflows, maintaining mapping consistency while supporting multiple tools.
3Device complexity
If simple heuristic rules are applied for automatic adjustment, then the algorithm is easier to implement and faster to execute, but it fails to address complex adjustments depending on scene characteristics
Solution Approach 1:
The patent replaces mechanical rule-based systems with a machine learning-based system. Instead of implementing complex if-then rules for different scene characteristics, the system uses a trained neural network that automatically learns to handle low key, high key, back-lighting, and other difficult situations through pattern recognition from training data.
Solution Approach 2:
The patent performs preliminary training of the machine learning model on extensive datasets of photographer adjustments before deployment. This preliminary action allows the system to pre-learn complex decision patterns for various scene characteristics, enabling it to handle complex adjustments accurately without requiring complex runtime logic.
Data Source
AI summary
Techniques are disclosed relating to generating generic labels, translating generic labels to image pipeline-specific labels, and automatically adjusting images. In one embodiment, generic labels may be generated. Generic algorithm parameters may be generated based on training a regression algorithm with the generic labels. The generic labels may be translated to pipeline-specific labels, which may be usable to automatically adjust an image.