Facial Image Superimposition for Low Light Clarity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image capturing apparatuses struggle to capture clear facial images when the subject is moving, especially in low light conditions, as extending exposure time leads to blur and existing correction techniques are ineffective for moving subjects.
Innovation Solution
An image capturing support apparatus with a detection unit to identify facial images, a determination unit to assess light sufficiency, and a superimposition unit that adjusts and combines facial image sizes to improve light conditions and reduce noise, enabling clear facial image capture.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Illumination intensity
If exposure time is extended to obtain sufficient light, then the amount of light for capturing facial images is improved, but the captured images become blurred when the subject is moving
Solution Approach 1:
The patent merges multiple captured images by detecting facial regions in each image, aligning them based on facial position, and superimposing them. This combination allows the system to use multiple short-exposure images to achieve the light-gathering effect of a long exposure while avoiding motion blur, as each individual image is captured with a short exposure time.
Solution Approach 2:
The patent dynamically adjusts the exposure time based on the movement state of the subject. When motion is detected, the system switches to a shorter exposure time to freeze motion, and then combines multiple such images to achieve sufficient brightness. This dynamic adaptation allows the system to optimize between light capture and motion freeze based on real-time conditions.
2Manufacturing precision
If multiple images are superimposed to correct blur, then image clarity is improved for stationary subjects, but the technique fails when the subject is moving
Solution Approach 1:
The patent applies local quality by focusing the superimposition operation specifically on the facial region rather than the entire image. By detecting facial regions in each image and aligning based on facial landmarks, the system ensures that the superimposition effectively combines facial details while being robust to overall image motion. This localized approach makes the technique adaptable to moving subjects.
3Productivity
If facial recognition is performed on blurred images, then the processing speed is maintained, but the recognition accuracy deteriorates
Solution Approach 1:
The patent performs preliminary image processing by superimposing multiple captured images to create a clear composite image before performing facial recognition. This preliminary action of creating a high-quality input image ensures that subsequent recognition algorithms work with optimal data, improving accuracy without significantly impacting overall processing speed since the superimposition operates on already-captured images.
Data Source
AI summary
An image capturing support apparatus 1 includes: a detection unit 2 configured to detect facial images that have facial areas corresponding to a face, from a plurality of captured images, respectively; a determination unit 3 configured to determine whether or not the amount of light is insufficient for the detected facial images, using the detected facial images; and a superimposition unit 4 configured to, when the amount of light is insufficient, adjust the size of each of the facial images, and superimpose the facial areas of the adjusted facial images on each other.