Single-Camera Face Image Extraction for Conference Emotion Reading
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In teleconferences, it is difficult for participants to read the facial expressions and emotions of multiple conference participants due to small face images captured by a single camera, hindering smooth communication.
Innovation Solution
An image processing apparatus and method that enhance facial expression detection and estimation, allowing for intuitive reading of feelings by changing display modes such as animation, color, and position based on estimated facial expressions and emotions, and reorganizing face images into a unified display.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Area of stationary object
If a single camera captures multiple conference participants, then the coverage of all participants is achieved, but the face image size becomes too small for reading facial expressions
Solution Approach 1:
The patent segments the wide-angle captured image by detecting individual face regions and extracting them as separate face images. This segmentation allows each face to be processed and displayed at an appropriate size while maintaining coverage of multiple participants.
Solution Approach 2:
The patent transitions from a two-dimensional wide-angle view to a multi-dimensional display by extracting faces in different directions (front, left, right) and arranging them in a three-dimensional-like layout on the screen, enabling larger face images while maintaining comprehensive coverage.
2Measurement precision
If facial expressions are enhanced through processing, then readability of emotions is improved, but processing complexity increases
Solution Approach 1:
The patent extracts only the essential facial region from the entire image, isolating the face area for specialized processing. This extraction reduces the complexity by focusing computational resources only on the relevant facial features rather than processing the entire image.
Solution Approach 2:
The patent performs preliminary face detection and region extraction before detailed expression analysis. By pre-identifying and isolating face regions, the system simplifies subsequent expression recognition tasks and reduces overall processing complexity.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
An image processing apparatus (10) includes an image data obtaining portion (100) that obtains image data of an image that captures a plurality of conference participants, a face image detecting portion (101) that detects a face image of each of the plurality of conference participants from the image data obtained by the image data obtaining portion, an image organizing portion (102) that extracts a detected face image and reorganizes detected face images into one image, a feeling estimating portion (104) that estimates a feeling of each conference participant based on the detected face image, and a display mode changing portion (105) that changes a display mode of the face image of each conference participant based on the estimated feeling.