Lens Distortion Correction via Segmented Horizontal and Vertical Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for correcting lens distortion in digital images are costly and inefficient, requiring large numbers of row buffers or inefficient DDR memory access.
Innovation Solution
A method and apparatus that correct lens distortion in two steps: first, correcting image data in a horizontal direction without the need for row buffers, and second, correcting in a vertical direction using dynamic memory with consecutive reading and writing, potentially with a 90-degree rotation to optimize DDR block data access.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If one-off lens distortion correction is implemented using an online row buffer, then lens distortion can be corrected in horizontal and vertical directions simultaneously, but a large number of row buffers are required leading to high cost
Solution Approach 1:
The patent divides lens distortion correction into two separate stages: first correcting horizontal distortion, then correcting vertical distortion. This segmentation allows each correction operation to use minimal row buffers (only 1-2 rows) instead of requiring large buffers to store entire distorted lines, thereby reducing device complexity while maintaining correction quality
Solution Approach 2:
The patent performs preliminary horizontal correction before vertical correction. By pre-correcting the horizontal distortion first, the subsequent vertical correction operates on already partially corrected data, reducing the buffer requirements for the second correction pass and overall system complexity
2Ease of manufacture
If one-off lens distortion correction is implemented using offline DDR, then correction can be performed after image data is stored, but DDR read efficiency is low due to nonconsecutive read addresses
Solution Approach 1:
The patent segments the correction process into horizontal and vertical passes, where each pass reads data sequentially in one direction. This segmentation transforms the original two-dimensional nonconsecutive access pattern into two separate one-dimensional sequential access patterns, dramatically improving DDR read efficiency while maintaining implementation flexibility
Solution Approach 2:
Instead of attempting to correct both horizontal and vertical distortion simultaneously (which requires nonconsecutive addressing), the patent inverts the approach by performing corrections sequentially in opposite directions. This inversion allows the DDR to read data in consecutive addresses during each pass, maximizing read efficiency
Data Source
AI summary
The present invention discloses a method and an apparatus for correcting lens distortion. In some feasible implementation manners of the present invention, image data acquired from a lens is corrected in a horizontal direction; the image data corrected in the horizontal direction is written into a dynamic memory; image data in the dynamic memory is corrected in a vertical direction; and lens distortion correction is performed in two steps: in the horizontal direction and in the vertical direction, so that lens distortion correction is implemented at a low cost. The present invention further discloses a corresponding apparatus for correcting lens distortion.