Image Processing With Feature-Point Alignment for Dynamic Clarity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image capturing technologies struggle to balance capturing a moving subject with dynamism while maintaining clarity and desirable facial expressions, especially when using fixed-point cameras or requiring high skill levels for panning shots, and existing multi-camera systems are costly and inefficient.
Innovation Solution
An image processing apparatus and method that detects subject areas and feature points in multiple images, aligning and synthesizing partial areas to match feature points, allowing for dynamic images with clear subjects and backgrounds, and optionally selecting images based on facial expressions or motion.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If multiple cameras are used to capture images with clear facial expressions, then the quality of captured images improves, but the device complexity and installation space requirements increase
Solution Approach 1:
The patent divides the image processing task into segments by detecting specific feature points (eyes, nose, mouth) and processing different regions of interest separately. This allows single-camera systems to achieve multi-camera quality by focusing computational resources on key facial areas rather than requiring multiple physical cameras to cover all aspects.
Solution Approach 2:
The patent creates virtual copies of facial features through image synthesis and processing. By detecting feature points and synthesizing images based on these points, the system replicates the effect of multiple cameras capturing different angles and expressions, achieving the same result with a single physical camera.
2Manufacturing precision
If fast shutter speed is used to capture moving subjects, then the clarity of the subject improves, but the sense of dynamism is lost
Solution Approach 1:
The patent applies dynamic processing by detecting feature points across multiple images captured at different moments and synthesizing them into a single image that preserves motion characteristics. The system dynamically adjusts the synthesis based on detected feature point positions, maintaining both subject clarity and the sense of movement through computational methods rather than static single-exposure capture.
3Object-generated harmful factors
If panning shooting is performed to capture dynamism, then the sense of motion improves, but the skill requirement and difficulty of operation increase
Solution Approach 1:
The patent enables the camera system to automatically perform the complex task of panning shot synthesis without requiring operator skill. The system self-manages the entire process: detecting subjects, tracking feature points across frames, determining optimal synthesis parameters, and generating the final image. This automated self-service approach eliminates the need for photographer skill in executing complex panning techniques while preserving the dynamic effect.
4Object-generated harmful factors
If long exposure period is used to capture subject motion, then the sense of dynamism improves, but the subject may move differently from predicted motion reducing accuracy
Solution Approach 1:
The patent uses periodic sampling of images at regular intervals during the exposure period, detecting feature points at each interval. This periodic detection approach allows the system to track actual subject motion patterns rather than relying on continuous prediction, capturing the subject's real movement trajectory and using these discrete samples to guide the synthesis process for maintaining accuracy while expressing dynamism.
Data Source
AI summary
An image processing apparatus comprises: a first detector that detects a subject area of a predetermined subject from each of a plurality of images; a second detector that detects a partial area of a predetermined size that includes the subject area from each of the plurality of images; a feature point detector that detects a feature point of an image; and a synthesizer that synthesizes the partial areas of the plurality of images so that the feature points in the subject areas match each other.