Abstract
Raw panoramas stitched from multiple images can have irregular shapes along their boundaries. Cropping of invalid pixels along the boundary is performed to generate a final result that is of a target shape. However, this can result in cropping of many valid pixels, reducing the overall portion of the captured scene depicted in the final result. This disclosure describes techniques to determine optimal content-preserving crops along the boundaries of a panorama in a manner that also optimizes latency and preserves aspect ratio. Per the techniques, in a first stage, dynamic programming is applied on predefined tiles to determine the largest rectangle on an initial stitched result. In a second stage, the rectangle is expanded based on rule-based configurable policies to further increase final pixel counts. In this manner, a two-stage boundary computation optimizes the number of valid pixels along the boundaries of a panorama.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Yu, Miao and Chu, Lun-Cheng, "Content-preserving Cropping for Panoramic Stitching of Images", Technical Disclosure Commons, (July 25, 2025)
https://www.tdcommons.org/dpubs_series/8393