Convolutional variational autoencoders for secure lossy image compression in remote sensing

Alessandro Giuliano,S. Andrew Gadsden,Waleed Hilal,John Yawney
2024-04-17
Abstract:The volume of remote sensing data is experiencing rapid growth, primarily due to the plethora of space and air platforms equipped with an array of sensors. Due to limited hardware and battery constraints the data is transmitted back to Earth for processing. The large amounts of data along with security concerns call for new compression and encryption techniques capable of preserving reconstruction quality while minimizing the transmission cost of this data back to Earth. This study investigates image compression based on convolutional variational autoencoders (CVAE), which are capable of substantially reducing the volume of transmitted data while guaranteeing secure lossy image reconstruction. CVAEs have been demonstrated to outperform conventional compression methods such as JPEG2000 by a substantial margin on compression benchmark datasets. The proposed model draws on the strength of the CVAEs capability to abstract data into highly insightful latent spaces, and combining it with the utilization of an entropy bottleneck is capable of finding an optimal balance between compressibility and reconstruction quality. The balance is reached by optimizing over a composite loss function that represents the rate-distortion curve.
Image and Video Processing,Machine Learning
What problem does this paper attempt to address?
The paper attempts to address the problem of achieving efficient and secure image compression in remote sensing data. Specifically, due to the rapid growth of remote sensing data and the limitations of hardware and battery that necessitate data transmission back to Earth for processing, new compression and encryption techniques are needed to minimize data transmission costs while ensuring reconstruction quality. The paper proposes a method based on Convolutional Variational Autoencoder (CVAE) for secure lossy image compression. This method can significantly reduce the amount of transmitted data while ensuring the security of image reconstruction and finding the optimal balance between compressibility and reconstruction quality by optimizing a composite loss function. The main contribution of the paper is to demonstrate the effectiveness of CVAE in remote sensing image compression, particularly in handling large-scale sensor images captured by small satellites (such as CubeSats), effectively compressing these images into smaller data products for more efficient transmission under bandwidth and power constraints. Additionally, the paper explores the specific implementation of the CVAE architecture and its performance under different dimensions, and experimentally verifies its ability to achieve significant compression ratios while maintaining high reconstruction quality.