A compact mathematical representation of human body silhouettes from frontal and lateral views

Document Type

Article

Publication Date

1-1-2024

Abstract

Human body silhouettes are used extensively in three-dimensional body shape modelling, activity recognition, apparel design, obesity, and posture assessment. These applications require efficient storage of human body images for future use and comparison. We proposed a novel one-dimensional mathematical representation of human body silhouettes from frontal and lateral views using a discrete cosine transform. Our method saved 75% of the storage space, significantly reducing costs, and achieved a compression ratio of 4:1 with an average reconstruction accuracy of 90% for all views of male and female images. Additionally, segment-wise silhouette representation decreased the average reconstruction complexity four times. Human body silhouettes are also modelled for the first time using polynomial curve fitting, discrete wavelet transform, and discrete Fourier transform with a systematic comparison. The polynomial curve fitting achieved the highest average space saving of 84%; however, reconstruction accuracy decreased by 12% compared to the discrete cosine transform. In addition, our novel method attained 46% additional storage space saving compared to standard two-dimensional JPEG and PNG image compression methods. Our work can be used to assess human body fat distribution, detect pose abnormalities and classify body shapes, ages, and genders.

Publication Source (Journal or Book title)

International Journal of Biomedical Engineering and Technology

First Page

102

Last Page

128

This document is currently not available here.

Share

COinS