Image Zone Processing with Singular Pixel Separation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing methods in embedded systems, such as digital cameras, often misclassify pixels, leading to visible artifacts due to incorrect classification of flat and textured zones, which can result in inappropriate noise reduction, sharpness enhancement, and contrast balancing.
Innovation Solution
A method that differentiates pixels into flat and textured zones by using a preliminary separation step to identify singular pixels, which are then processed separately to avoid misclassification, and employs adaptive noise reduction, sharpness increase, and acutance improvement using unsharp masks, while optimizing memory resources by reusing reference kernels and processing in parallel.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a simple classification criterion is used to differentiate flat and textured zones, then the processing complexity is reduced, but misclassification of pixels occurs leading to visible artifacts
Solution Approach 1:
The patent applies preliminary action by performing a first classification of pixels into singular and non-singular categories before the main flat/textured zone classification. This preliminary step identifies pixels with unusual characteristics (such as isolated different values in the kernel) that would likely be misclassified, allowing them to be processed separately and preventing the propagation of classification errors through the image processing pipeline.
2Manufacturing precision
If multiple image processing processes are applied sequentially, then the visual quality is improved, but the memory resources are excessively consumed
Solution Approach 1:
The patent merges the reference kernels used in sequential processing steps by reusing the same kernel structure and data across noise reduction, sharpness enhancement, and acutance improvement processes. Instead of creating separate kernel copies for each process, the system maintains a single reference kernel in memory that is updated and reused throughout the processing sequence, significantly reducing memory consumption while preserving the ability to apply multiple enhancement processes.
Data Source
AI summary
A method for improving the perception of an image may include performing a main separation of the pixels of the image into two categories, one corresponding to pixels of a flat zone, and the other corresponding to pixels of a textured zone. The method may also include processing the pixels of each category according to a method optimized according to the type of zone. Before the main separation step, a preliminary separation of the pixels may be performed into one category of normal pixels intended for the main separation step, and one category of singular pixels, with the criterion for selecting the singular pixels being adapted to identify pixels that would be wrongly identified as pixels of a textured zone. The singular pixels may then be processed according to a method adapted to their nature.