Polygon Filtering for Image Boundary Jitter Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Image processing technologies in digital devices struggle to accurately determine the boundary of a target object, leading to jitter and boundary inaccuracy due to inherent imaging variations and user movements, which affects the user experience in document capture and sharing applications.
Innovation Solution
The implementation of intra- and inter-frame polygon filtering techniques to analyze and stabilize the candidate object boundaries by applying filters based on area and position variance, ensuring a statistically high certainty in identifying the target object boundary.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If native image processing capabilities are used to detect polygons in video frames, then the boundary detection process is simple and rapid, but the detected boundary exhibits jitter and instability due to hand or camera movements and imaging variations
Solution Approach 1:
The system performs preliminary actions by detecting all polygons in each frame and storing them before final boundary determination. This allows the system to have candidate boundaries ready in advance, which are then evaluated using stability criteria to select the most reliable boundary, thus resolving the contradiction between rapid detection and stable results
Solution Approach 2:
The system implements feedback by evaluating the stability of detected polygons across multiple frames and using this information to adjust boundary selection. The stability evaluation provides feedback on which detected boundaries are reliable, allowing the system to prefer stable boundaries over unstable ones, thereby improving boundary reliability while maintaining detection speed
2Device complexity
If the largest detected polygon is used to define the target document boundary, then the boundary determination is computationally simple, but the boundary may be inaccurate when multiple objects or shadows are present in the frame
Solution Approach 1:
The system introduces an intermediary stability evaluation mechanism between polygon detection and boundary selection. Instead of directly using the largest polygon, the system evaluates the stability of candidate polygons across frames and uses this intermediate assessment to select the most accurate boundary, effectively filtering out false positives from shadows or other objects
Solution Approach 2:
The system changes the selection parameter from simply 'largest area' to 'most stable across frames'. This parameter change allows the system to distinguish between actual document boundaries and spurious detections from shadows or other objects, improving boundary accuracy while maintaining computational simplicity through the use of clear selection criteria
3Speed
If polygon detection is performed at 30 frames per second, then the boundary detection is continuous and responsive, but the temporal inter-frame uncertainty causes visible jitter in the displayed boundary
Solution Approach 1:
The system performs preliminary detection of all candidate polygons at 30 fps and stores them for later evaluation. This allows continuous detection to maintain responsiveness while the subsequent stability-based selection process ensures temporal consistency, resolving the jitter caused by frame-to-frame variations
Solution Approach 2:
Instead of selecting the first or largest polygon detected, the system inverts the approach by selecting the polygon that shows the most consistent position and size across multiple frames. This inverted selection criterion prioritizes temporal consistency over detection order or size, effectively eliminating visible jitter while maintaining 30 fps detection rate
Data Source
AI summary
Image processing to reduce or eliminate jitter when visually highlighting a target capture area of an image capture device such as a smart phone or camera. A method embodiment commences upon receiving a sequence of one or more video frames taken by the image capture device. A routine to identify the largest polygons in the video frames is applied to the sequence. Uncertainty in identification of the largest polygons is reduced or eliminated by applying intra-frame filters to the polygons. Visible jitter that can arise from inter-frame uncertainty when selecting and highlighting target capture areas can be reduced or eliminated by applying inter-frame filters so as to retain selection of a particular target capture area even in the presence of certain capture device movements. A jitter-free representation of the target capture area is visually displayed on a display screen of the image capture device.