TD-dAJC: A 2pJ/Pixel Time-Domain Weight and Integrating-MAC Based Direct-Analog-to-MJPEG Compression for Video Sensor Nodes
The increase in IoT applications has made cameras ubiquitous in various fields spanning from healthcare and autonomous vehicles to energy-constrained applications like battery-powered wearables. This proliferation has created a need for low-power camera modules for resource-constrained IoT nodes/edge devices. As the quality of images improves, these cameras produce a data deluge. Standard digital cameras generate data in the order of ~Gbps, which requires compressing the data within the camera module (e.g., SONY IMX 317) using lossy compression schemes like MJPEG or H.264. For instance, a 4K 12fps video generates around ~1 Gbps of data, which can be compressed to ~50 Mbps using MJPEG. Compared to H.264, MJPEG has lower latency and lower complexity, making it suitable for real-time edge applications. Standard digital data compression suffers from 1) high ADC energy consumption digitizing all samples, 2) Intermediate data storage requirements, and 3) huge computation power to compress the data, making it less desirable for edge cameras. The emergence of low-power camera imagers [1], such as the STM VD55G1 with power consumption as low as 3-10mW, has increased the demand for efficient compression techniques that minimize computational power as shown in Fig. 1(a) and (b).
Authors
- Shreyas Sen (ORCID: https://orcid.org/0000-0001-5566-8946)
- Gourab Barik (ORCID: https://orcid.org/0009-0006-7556-7273)
- Yudhajit Ray (ORCID: https://orcid.org/0000-0001-9824-3788)
- Harshit Naman
Institutions
- Purdue University West Lafayette (US)
Publication Details
- Published
- 2025-04-13
- DOI
- https://doi.org/10.1109/cicc63670.2025.10982732
- Citations
- 1
- Primary Topic
- Sparse and Compressive Sensing Techniques
- Type
- article
- Field-Weighted Citation Impact
- 1.33