Image Processing Device for Document Inclination Correction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing techniques fail to accurately correct the inclination of document images during scanning, leading to inappropriate output images, especially when dealing with non-rectangular documents like receipts.
Innovation Solution
An image processing device and method that detects two straight lines within the scanned image, calculates evaluation values based on their relationships with the document edges, and uses these values to select the appropriate straight lines for rotating the image data to correct the inclination, thereby generating accurate output images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If baseline detection method is used to correct image inclination, then measurement precision may be improved, but device complexity and processing time increase significantly
Solution Approach 1:
The patent extracts only the essential straight lines (document edges) needed for inclination correction, rather than performing comprehensive baseline detection. By detecting straight lines and selecting those with the longest length that correspond to document edges, the method obtains sufficient information for correction without the complexity of full baseline analysis.
Solution Approach 2:
The patent segments the image processing into distinct stages: straight line detection, line selection based on length and orientation criteria, and inclination correction. This segmentation allows the system to achieve necessary precision through simplified sub-tasks rather than attempting comprehensive baseline detection in one complex operation.
2Measurement precision
If comprehensive baseline detection is performed, then inclination correction accuracy improves, but processing time increases
Solution Approach 1:
The patent performs partial action by detecting only the most prominent straight lines (those with longest length) rather than analyzing all possible baselines. By selecting straight lines that meet specific length and orientation criteria, the method achieves sufficient correction accuracy with reduced processing time compared to comprehensive baseline detection.
Solution Approach 2:
The patent changes the detection parameters from comprehensive baseline analysis to straight line detection with length-based filtering. By setting minimum length thresholds and orientation constraints, the system efficiently identifies the relevant document edges for correction without exhaustive search, significantly reducing processing time while maintaining accuracy.
3Productivity
If simple straight line detection is used, then processing speed improves, but reliability of inclination correction deteriorates
Solution Approach 1:
The patent incorporates feedback mechanisms through evaluation values that assess the suitability of detected straight lines. The system calculates evaluation values based on line length, orientation, and positional relationships, then uses this feedback to select the most appropriate straight lines for correction, ensuring reliable results while maintaining processing efficiency.
Solution Approach 2:
The patent enhances simple straight line detection by introducing multiple selection criteria: minimum length thresholds, orientation constraints (horizontal or vertical), and evaluation value comparisons. These parameter changes transform basic detection into a robust selection process that ensures correction reliability without sacrificing processing speed.
4Measurement precision
If multiple straight lines are detected and evaluated, then selection accuracy improves, but device complexity increases
Solution Approach 1:
The patent applies local quality by treating different straight lines with different evaluation criteria based on their characteristics. Rather than using a uniform complex algorithm for all lines, the system evaluates each line based on its specific properties (length, orientation, position) and selects those that best match the document edge characteristics, achieving high selection accuracy with manageable complexity.
Data Source
AI summary
An image processing device may perform: obtaining first original image data generated by scanning a document including an object, the first original image data representing a first original image which includes a first object image indicating the object; detecting two or more straight lines in the first original image; specifying a first relationship between a first straight line among the two or more straight lines and the first object image; selecting, based on the first relationship, a particular straight line among the two or more straight lines; outputting, based on a direction in which the particular straight line extends, an output image represented by output data, the output data being generated by using the first original image data.