Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2024 Sep 4;24(17):5759.
doi: 10.3390/s24175759.

Local-Peak Scale-Invariant Feature Transform for Fast and Random Image Stitching

Hao Li  1 , Lipo Wang  2 , Tianyun Zhao  3 , Wei Zhao  1
Affiliations

Local-Peak Scale-Invariant Feature Transform for Fast and Random Image Stitching

Hao Li et al. Sensors (Basel). .

Abstract

Image stitching aims to construct a wide field of view with high spatial resolution, which cannot be achieved in a single exposure. Typically, conventional image stitching techniques, other than deep learning, require complex computation and are thus computationally expensive, especially for stitching large raw images. In this study, inspired by the multiscale feature of fluid turbulence, we developed a fast feature point detection algorithm named local-peak scale-invariant feature transform (LP-SIFT), based on the multiscale local peaks and scale-invariant feature transform method. By combining LP-SIFT and RANSAC in image stitching, the stitching speed can be improved by orders compared with the original SIFT method. Benefiting from the adjustable size of the interrogation window, the LP-SIFT algorithm demonstrates comparable or even less stitching time than the other commonly used algorithms, while achieving comparable or even better stitching results. Nine large images (over 2600 × 1600 pixels), arranged randomly without prior knowledge, can be stitched within 158.94 s. The algorithm is highly practical for applications requiring a wide field of view in diverse application scenes, e.g., terrain mapping, biological analysis, and even criminal investigation.

Keywords: LP-SIFT; image mosaic; image stitching.

PubMed Disclaimer

Conflict of interest statement

The authors declare no conflicts of interest.

Figures

Figure 1
Figure 1
Framework of the LP-SIFT method.
Figure 2
Figure 2
Diagram of the LP-SIFT method.
Figure 3
Figure 3
Datasets. Dataset-A: (a) mountain [52] dataset image pair, (b) street view [53] dataset image pair, (c) terrain [54] dataset image pair. Dataset-B: (d) building dataset image pair, (e) campus view dataset (translation) image pair, (f) campus view dataset (rotation) image pair.
Figure 4
Figure 4
Stitching results of mountain dataset and street dataset. (a) Mountain dataset stitched by SIFT, ORB, BRISK, SURF, and LP-SIFT, respectively. In LP-SIFT, L = [32, 40]. (b) Street view dataset stitched by SIFT, ORB, BRISK, SURF, and LP-SIFT, respectively. In LP-SIFT, L = [32, 40].
Figure 5
Figure 5
Comparison of the stitching times of 5 algorithms for different datasets.
Figure 6
Figure 6
Stitching results of terrain dataset and building dataset. (a) Terrain dataset stitched by SIFT, ORB, BRISK, SURF, and LP-SIFT respectively. In LP-SIFT, L = [32,64]. (b) Building dataset stitched by ORB, BRISK, SURF, and LP-SIFT respectively. In LP-SIFT, L = [100,128].
Figure 7
Figure 7
Stitching results of campus view dataset. (a) Campus view (translation) dataset stitched by BRISK, SURF, and LP-SIFT, respectively. In LP-SIFT, L = [256,512]. (b) Campus view (rotation) dataset stitched by SURF, and LP-SIFT, respectively. In LP-SIFT, L = [256,512].
Figure 8
Figure 8
Schematic diagram of LP-SIFT image mosaic of multiple images without prior knowledge.
Figure 9
Figure 9
Mosaic of multiple images by LP-SIFT without prior knowledge, where L = [512,1024]. (a) Original image; the image size is 6400 × 4270. (b) The original image is stitched into different sizes and its position is shuffled, and its size is marked below the image. (c) Stitching result, and the stitching time is 158.94 s.
See this image and copyright information in PMC

References

    1. Deshmukh P., Paikrao P. A Review of Various Image Mosaicing Techniques; Proceedings of the 2019 Innovations in Power and Advanced Computing Technologies (i-PACT); Vellore, India. 22–23 March 2019; - DOI
    1. Lee W.T., Chen H.I., Chen M.S., Shen I.C., Chen B.Y. High-resolution 360 Video Foveated Stitching for Real-Time VR. Comput. Graph. Forum. 2017;36:115–123. doi: 10.1111/cgf.13277. - DOI
    1. Saeed S., Kakli M.U., Cho Y., Seo J., Park U. A High-Quality Vr Calibration and Real-Time Stitching Framework Using Preprocessed Features. IEEE Access. 2020;8:190300–190311. doi: 10.1109/ACCESS.2020.3031413. - DOI
    1. Dang P., Zhu J., Zhou Y., Rao Y., You J., Wu J., Zhang M., Li W. A 3D-Panoramic Fusion Flood Enhanced Visualization Method for VR. Environ. Model. Softw. 2023;169:105810. doi: 10.1016/j.envsoft.2023.105810. - DOI
    1. Liu Z., Chang S. A Study of Digital Exhibition Visual Design Led by Digital Twin and VR Technology. Meas. Sens. 2024;31:100970. doi: 10.1016/j.measen.2023.100970. - DOI

LinkOut - more resources