Facial Feature Localization Using Iterative Active Shape Model
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for identifying facial features in video images are inefficient, requiring high computational power and causing discontinuous motion of facial features during video display, leading to unsatisfactory performance in electronic equipment.
Innovation Solution
A method involving image tracing, data calculation, and iterative localization of facial features using a video tracing technique and active shape model, which calculates video data differences and sets iteration numbers to improve processing speed and display performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If scanning each video image to identify facial features is performed, then facial feature identification accuracy is improved, but processing speed deteriorates due to huge data workloads
Solution Approach 1:
The patent applies preliminary action by using the facial feature coordinates from the previous frame as initial values for the current frame identification. This pre-positioning of coordinates based on temporal continuity reduces the search space and computational burden, thereby improving processing speed while maintaining identification accuracy through the active shape model iterative optimization.
2Device complexity
If facial features are identified without considering relative face size, then processing simplicity is improved, but motion smoothness deteriorates during video display
Solution Approach 1:
The patent applies parameter changes by dynamically adjusting the active shape model parameters based on the relative size of the face in each frame. This allows the model to adapt to scale variations while maintaining consistent facial feature localization, ensuring smooth motion during video display without requiring complex processing of absolute coordinate transformations.
3Measurement precision
If traditional facial feature identification method is used, then comprehensive feature detection is improved, but computational resource consumption increases
Solution Approach 1:
The patent applies segmentation by dividing the facial feature identification process into two stages: first using the active shape model to locate key facial landmarks, then using these landmarks to define regions of interest for more detailed feature analysis. This segmented approach ensures comprehensive feature detection while reducing overall computational resource consumption by focusing intensive processing only where needed.
Data Source
AI summary
A method for identifying facial features comprises following steps: An image tracing step is performed to receive video data of a plurality of face images and to obtain a real-time background image from the video data by a video tracing technique. A data calculating step is performed to calculate a video data difference between a current face image and the real-time background image. A process setting step is performed to set an iteration number according to the video data difference. A coordinate requesting step is performed to obtain facial feature coordinates of a previous face image, the previous one of the current face image, serving as initial facial feature coordinates. A localization step is performed to obtain current facial feature coordinates of the current image by conducting an iterative calculation according to the iteration number and based on the initial facial feature coordinates.