Text Recognition in Video Streams Using Per-Character Weighting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing text recognition systems face challenges in accurately recognizing text under conditions of projective distortions and low lighting, particularly in uncontrolled environments such as those encountered when using mobile devices to capture documents like identity documents.
Innovation Solution
A method involving the use of multiple image frames from a video stream, where text recognition results are combined using per-character weighting, with frame weights and character weights calculated using a focus estimation algorithm to improve recognition accuracy by aligning and accumulating character recognition results across frames.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If text recognition is performed on a single image frame, then the processing speed is fast, but the recognition accuracy deteriorates under projective distortions and low lighting conditions
Solution Approach 1:
The patent segments the text recognition task into per-character processing units. Each character is independently recognized and weighted based on local image quality metrics, allowing the system to handle projective distortions and low lighting conditions at the character level while maintaining overall processing efficiency.
Solution Approach 2:
The patent applies partial action by selectively processing only the most reliable character recognitions first, using confidence thresholds and quality metrics to determine which characters require reprocessing or alternative handling methods, thereby reducing unnecessary computational overhead.
2Measurement precision
If multiple image frames are combined to improve recognition accuracy, then the text recognition accuracy improves, but the processing time increases
Solution Approach 1:
The patent performs preliminary quality assessment on each image frame before full processing, using focus estimation and lighting condition analysis to pre-rank frames. This allows the system to prioritize high-quality frames for detailed processing while quickly discarding or minimally processing low-quality frames, reducing overall processing time.
Solution Approach 2:
The patent dynamically adjusts processing parameters such as the number of frames to process, confidence thresholds, and quality metric weights based on the specific characteristics of the input video stream, optimizing the balance between accuracy and processing speed for different scenarios.
3Measurement precision
If per-character weighting is applied to combine recognition results, then the recognition accuracy under distortions improves, but the computational complexity increases
Solution Approach 1:
The patent applies local quality assessment by calculating image quality metrics (focus, lighting, distortion) specifically for each character's local region rather than for the entire image. This allows the system to weight individual character recognitions based on their local conditions, improving accuracy while keeping computational complexity manageable through localized processing.
4Measurement precision
If all image frames are processed to maximize accuracy, then the recognition accuracy improves, but the resource consumption increases
Solution Approach 1:
The patent implements partial processing by using confidence thresholds and quality metrics to identify characters and frames that meet minimum reliability criteria. Only those requiring improvement are subjected to additional processing steps, while sufficiently reliable recognitions are accepted immediately, reducing energy consumption while maintaining accuracy.
Data Source
AI summary
Text recognition in a video stream using combined recognition results with per-character weighting. In an embodiment, for each frame in a video stream, a text-recognition result is obtained and a frame weight is calculated. The text-recognition results of the frames are combined by aligning character-recognition results and calculating a character weight for each character-recognition result. At each position in the alignment, the character-recognition results are accumulated based on the character weights and frame weights to produce an accumulated text-recognition result that represents a text field in the video stream.