Related Experiment Video
Updated: May 25, 2026
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
Gaussian multiscale aggregation applied to segmentation in hand biometrics.
Alberto de Santos Sierra1, Carmen Sánchez Avila, Javier Guerra Casanova
1Group of Biometrics, Biosignals and Security, Universidad Politécnica de Madrid, Campus de Montegancedo s/n, 28223 Pozuelo de Alarcón, Madrid, Spain. alberto@cedint.upm.es
Sensors (Basel, Switzerland)
|January 17, 2012
Summary
A new image segmentation algorithm effectively isolates hands from complex backgrounds. This method offers superior accuracy and efficiency for hand biometrics compared to existing techniques.
Area of Science:
- Computer Vision
- Biometrics
- Image Processing
Background:
- Accurate hand segmentation is crucial for biometric applications.
- Existing methods face challenges with diverse background textures.
Purpose of the Study:
- To introduce a novel image segmentation algorithm for hand biometrics.
- To evaluate its performance against established methods.
Main Methods:
- Gaussian multiscale aggregation for image segmentation.
- Utilized a large synthetic database of 408,000 hand images.
- Compared against Lossy Data Compression (LDC) and Normalized Cuts (NCuts).
Main Results:
- The proposed algorithm successfully segmented hands from various backgrounds (carpets, fabric, glass, grass, soil, stones).
- Outperformed LDC and NCuts in accuracy, computational cost, time performance, and memory usage.
- Demonstrated significant improvements in segmentation efficiency.
Conclusions:
- The Gaussian multiscale aggregation algorithm is a highly effective solution for hand segmentation in biometrics.
- Offers a computationally efficient and accurate alternative to current methods.
- Potential for widespread adoption in hand-based identification systems.
