Projective Transformation for Trapezoidal Distortion Correction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing panoramic image capturing techniques produce low-quality images when the image capturing apparatus is in a tilted posture, resulting in distorted and misaligned building images due to trapezoidal distortion, especially when capturing rectangular buildings.
Innovation Solution
An image processing apparatus that calculates the movement along the image capturing direction, performs projective transformation based on the posture of the image capturing apparatus, and aligns images to correct trapezoidal distortion, allowing for the generation of high-quality panoramic images by joining images in the correct order.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If panoramic image capturing is performed with the image capturing apparatus in a tilted posture (optical axis tilted upward or downward), then the capturing flexibility and ease of operation are improved, but the image quality deteriorates due to trapezoidal distortion and misalignment
Solution Approach 1:
The patent applies preliminary action by detecting the posture of the image capturing apparatus before image combining, and performing projective transformation to correct trapezoidal distortion in advance. The processor calculates the amount of movement along the image capturing direction based on detected posture information, and performs projective transformation on each image to correct distortion before alignment and combining, thereby preventing quality deterioration from occurring in the final panoramic image
Solution Approach 2:
The patent applies parameter changes by using detected posture information (tilt angle) to dynamically calculate the amount of movement for image alignment and to determine projective transformation parameters. The system changes the transformation parameters based on the actual posture parameters detected, allowing the image processing to adapt to different tilting conditions and maintain high image quality regardless of the capturing posture
2Adaptability or versatility
If images are captured while the image capturing direction is sequentially changed in a tilted posture, then the capturing versatility is improved, but the alignment precision deteriorates resulting in curved buildings and misaligned joints
Solution Approach 1:
The patent performs preliminary projective transformation to correct trapezoidal distortion based on detected posture information before performing image alignment. By calculating the amount of movement along the image capturing direction in advance and applying the appropriate transformation, the system ensures that buildings and other vertical structures remain straight and properly aligned in the final panoramic image, regardless of the tilting angle used during capturing
Solution Approach 2:
The patent employs feedback by using the detected posture information of the image capturing apparatus to dynamically adjust the image processing parameters. The system continuously monitors the tilt angle and uses this feedback to calculate the correct amount of movement for alignment and to apply appropriate projective transformation, ensuring accurate alignment precision adaptively responds to different capturing conditions
Data Source
AI summary
A image signal processing circuit acquires a plurality of images captured while an image capturing direction is sequentially changed. A microcomputer calculates the amount of movement along image capturing direction at the time when the plurality of images is sequentially captured, and calculates a projective transformation coefficient for each of the plurality of images based on a posture of an image capturing apparatus at the time when each of the plurality of images is captured. Then, the image signal processing circuit performs a projective transformation using the coefficient to correct trapezoidal distortions in the images, and, performs alignment between the images of neighboring image capturing direction after the projective transformation is performed, based on the amount of movement, and combines the images on which the alignment has been performed by joining the images together in the order in which the image capturing direction is changed.