Scrolling Text Region Detection Using Edge and Stability Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing methods struggle to accurately detect scrolling text regions in video sequences due to the lack of clear edge information and intricate backgrounds, leading to false positives and visual artifacts.
Innovation Solution
An image processing method involving generating refined edge maps, edge sum images, and intensity difference images to identify text and banner regions, and performing stability checks on their indices before enhancing the detected text regions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional text detection methods are used on video frames, then processing speed is maintained, but detection accuracy deteriorates due to lack of clear edge information and intricate backgrounds
Solution Approach 1:
The patent segments the text detection problem into multiple stages: edge detection to generate edge maps, edge map refinement to remove false edges, edge sum calculation to identify potential text regions, and intensity difference analysis to confirm text boundaries. This segmentation allows each stage to focus on specific features, improving overall detection accuracy while managing complexity through modular processing
Solution Approach 2:
The patent performs preliminary edge detection and refinement operations on reference frames before text detection. By pre-processing the edge maps and identifying potential text regions in advance, the system reduces the complexity of subsequent detection steps and improves accuracy by working with pre-processed, cleaned edge information rather than raw frame data
2Reliability
If text detection is performed on every frame, then detection reliability is improved, but processing time increases
Solution Approach 1:
The patent merges multiple frames to generate composite edge maps and intensity difference images. By combining information from multiple frames, the system improves detection reliability through accumulated evidence while reducing the need for intensive processing on every individual frame. The merging operation allows the system to identify stable text regions that persist across frames, reducing false positives without requiring full processing of each frame
Solution Approach 2:
The patent implements periodic stability checks on detected text regions by comparing detections across multiple frames. Instead of performing full detection on every frame, the system periodically verifies text region stability by checking consistency across frames, which maintains reliability while reducing overall processing time through selective verification rather than continuous full processing
3Manufacturing precision
If motion compensation is applied to scrolling text regions, then image quality is improved, but false positives increase due to inaccurate region detection
Solution Approach 1:
The patent implements feedback mechanisms where detected text regions from one frame inform the detection process in subsequent frames. By using previously detected text region locations and characteristics as feedback for future detections, the system improves precision in identifying scrolling text while reducing false positives through consistency checking. The feedback loop allows the system to learn from past detections and adjust subsequent detection parameters accordingly
Solution Approach 2:
The patent introduces edge sum images and intensity difference images as intermediary representations between raw frame data and final text detection results. These intermediary structures serve as mediators that filter and organize information, making it easier to identify true text regions while ignoring false edges. The edge sum operation acts as an intermediary that aggregates edge information in a way that highlights text regions while suppressing background noise, reducing false positives before motion compensation is applied
Data Source
AI summary
An image processing method is provided, including the following operations: generating, according to a refined edge map of a first frame in the video sequence, an edge sum image including row scores corresponding to rows in the refined edge map; generating first and second intensity difference images that indicate row differences in image data between adjacent rows of the rows; identifying, according to the edge sum image and the first to second intensity difference images, indices of a first text region and a first banner region that are in the first frame; determining whether the indices of the first text region and the first banner region in the first frame meet a stable condition; and performing image processing operation to the first text region when the indices of the first text region and the first banner region in the first frame meet the stable condition.