Related Experiment Videos
Seamless image stitching by minimizing false edges.
Assaf Zomet1, Anat Levin, Shmuel Peleg
1School of Computer Science and Engineering, The Hebrew University of Jerusalem, 91904 Jerusalem, Israel. zomet@humaneyes.com
Summary
This study introduces novel cost functions for image stitching, optimizing seam visibility and similarity to original images in the gradient domain. This approach enhances stitching quality for applications like panoramic image generation and object blending.
Area of Science:
- Computer Vision
- Image Processing
- Computational Photography
Background:
- Image stitching is crucial for applications like mosaicing and object insertion.
- Current stitching quality assessment relies on visual similarity and seam visibility.
- Photometric inconsistencies and geometric misalignments challenge stitching accuracy.
Purpose of the Study:
- To introduce formal cost functions for evaluating image stitching quality.
- To define similarity and seam visibility in the gradient domain for optimized stitching.
- To minimize disturbing edges along the seam for seamless image integration.
Main Methods:
- Development of several formal cost functions for stitching quality evaluation.
- Defining similarity and seam visibility in the gradient domain.
- Minimizing disturbing edges by optimizing cost functions.
Main Results:
- The proposed cost functions effectively evaluate stitching quality by considering gradient domain properties.
- Optimization of these functions addresses photometric and geometric inconsistencies.
- Demonstrated effectiveness in panoramic image generation, object blending, and artifact removal.
Conclusions:
- Optimizing image stitching in the gradient domain offers significant benefits over existing methods.
- The novel cost functions provide a robust framework for high-quality image stitching.
- This approach improves visual fidelity and reduces artifacts in stitched images.