Image Processing Device Lip Shape Feature Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing technologies struggle to quickly and accurately identify the meaning a speaker intends to convey from input images.
Innovation Solution
An image processing device that groups input images into bundles of a certain size, divides each image into patch images of a corresponding size, and derives feature data from lip shapes using these patch images, with a correction unit to refine the feature data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If input images are processed individually without grouping, then processing simplicity is maintained, but processing speed and accuracy are insufficient
Solution Approach 1:
The patent divides the image processing task into two segments: first grouping multiple input images into bundles, then processing each bundle as a unit. This segmentation enables parallel processing of multiple images simultaneously, improving productivity while maintaining manageable complexity through structured organization.
Solution Approach 2:
The patent merges multiple input images into bundles before processing. By combining multiple images into a single processing unit (bundle), the system can process multiple images concurrently, thereby increasing processing speed and productivity without treating each image completely independently.
2Measurement precision
If images are divided into smaller patches, then feature extraction accuracy improves, but processing time increases
Solution Approach 1:
The patent segments each input image into smaller patches (e.g., 16x16 or 32x32 pixel patches) to enable fine-grained feature extraction. This segmentation allows the system to capture detailed lip shape features accurately while the grouping mechanism ensures that multiple such detailed extractions can proceed in parallel.
3Measurement precision
If the number of processing units is increased, then feature data accuracy improves, but computational load increases
Solution Approach 1:
The patent combines multiple input images into bundles that are then processed together as a single unit. This merging approach allows the system to utilize multiple processing units efficiently by distributing the computation across bundles rather than processing each image separately, thereby improving accuracy without proportionally increasing computational load.
Data Source
AI summary
The image processing device according to the present disclosure may more quickly and accurately identify the meaning that a speaker intends to convey by grouping a plurality of input images into bundles of a group size corresponding to a certain size and then deriving feature data corresponding to lip shapes included in the plurality of input images based on a patch image obtained by dividing each of the plurality of input images into a patch size corresponding to the certain size.