Energy-Efficient Image Compression for Capsule Endoscopy Using a CNN-Based Feature Learning Algorithm
S Harshitha, Udigala Basavaraju Mahadevaswamy, Mallikarjunaswamy Srikantaswamy · Engineering Technology & Applied Science Research · 2025
Wireless Capsule Endoscopy (WCE) has revolutionized Gastrointestinal (GI) diagnostics by allowing non-invasive internal visualization. However, it generates massive amount of image data, leading to considerable memory and power requirements in terms of transmission and storage in battery-constrained applications. In WCE systems, conventional methods of image compression like Discrete Cosine Transform (DCT), Discrete Wavelet Transform (DWT), and Set Partitioning in Hierarchical Trees (SPIHT) are widely applied. Such algorithms, although efficient under typical conditions, have limitations such as increased computational complexity, poor power efficiency, and reduced image quality at elevated compression levels. To overcome these limitations, the present study proposes a new technique known as the CNN-based Feature Learning Compression Algorithm (CFLCA). This technique deploys Convolutional Neural Networks (CNNs) to obtain optimal spatial features for more efficient image compression in terms of energy consumption and memory usage. The model is trained to maintain a trade-off between image quality and compression ratio using Peak Signal-to-Noise Ratio (PSNR) as a metric. The experimental results demonstrate that the suggested CFLCA achieves a 0.28% improvement in compression ratio, a 0.15% increase in PSNR, and a 0.22% reduction in power consumption compared to traditional methods. These improvements show the promise of CFLCA in facilitating real-time and efficient image compression in energy-limited wireless medical imaging applications.