Compression, once considered a compromise in high-bandwidth machine vision, is now used extensively as video becomes an essential tool across more industries.

In terms of technology, most people first encounter compression through audio. Audiophiles can argue for hours about the poor fidelity of streaming music services versus physical media. In a previous role, our R&D team worked extensively on preserving audio quality while employing various compression techniques for voice over IP, cellular, in-car, and teleconference services.

In vision systems, H.264 compression is commonly applied at the sensor level in embedded applications where visual appearance takes precedence over data integrity. This lossy compression approach helps minimize power, footprint, storage and bandwidth demands for wireless transmission in consumer or commercial drone, robotics, and IP camera surveillance applications. Lossy compression is also beneficial in edge processing applications, where raw data is analyzed on-device and only metadata or events are transmitted to processing.

A lossy, or “visually lossless,” method produces video that is indistinguishable to the human eye. Designers can benefit from higher compression ratios, but removing or altering some data does compromise the quality and performance of processing. This makes it unsuitable for most real-time applications where data integrity is critical, such as real-time security monitoring and medical imaging.

Mathematically lossless compression, in comparison, ensures imaging data remains intact while meeting latency demands for real-time processing. From sensor to processor, data remains an exact bit-for-bit match, though at a lower compression ratio versus lossy techniques.

Lossless compression approaches include JPEG-LS and JPEG 2000, and a patented hardware acceleration technique. The technique compresses imaging data line-by-line with under two image lines of latency, making it ideal for real-time applications. The algorithm dynamically selects the best compression profile for each line and includes a feature allowing users to reduce CPU load during decompression.

Advantages of Mathematically Lossless Compression for Imaging

A mathematically lossless compression approach provides several key benefits for imaging device designers and integrators to meet growing demands for higher resolutions and low latency performance while managing total system costs.

For imaging devices, lossless compression enables the use of high-resolution sensors that generate large volumes of data that can be transmitted over limited networking infrastructure. This allows more images to be sent in the same time frame, or for multiple high-bandwidth sources to operate on existing networks. Many industries rely on legacy infrastructure not built for today’s data demands, and upgrading is expensive.

Compression and Perimeter Security

Secure facilities, including airports, military bases, borders, and commercial warehouses, are adopting more advanced imaging systems integrating infrared and visible capabilities to automatically detect and track potential threats. Infrared imaging consumes significant bandwidth, especially when uncompressed, and multicasting to multiple processing, display, and storage systems also increases network demands.

In one case, an integrator has developed an infrared 360° camera system for airport security. The system integrates multiple cameras to provide real-time detection, classification and tracking to safeguard against unlawful and accidental intrusions, protect critical areas, and manage the movement of personnel, aircraft, and vehicles.

Here, a video interface with integrated lossless compression transmits real-time, low-latency video from the camera over extended-reach Ethernet cabling to a central office. Power, control signals, and data are all transmitted over a single, thinner, lighter, and more flexible cable. With GigE Vision, multiple cameras are networked to a single host PC for processing and display and new cameras are easily added to the network. Video can also be multicast to processing, displays, and storage systems.

The lossless compression algorithm transmits up to 2 Gbps of video over 1 Gbps Ethernet, reducing infrastructure and system costs while lowering power consumption. This nearly doubles the effective bandwidth and allows additional sensors without overloading the network.

Compression nearly doubles the effective bandwidth of the existing infrastructure, enabling the system expansion with additional sensors without overloading the network.

Cost and Patient Advantages in X-ray Imaging

Mathematically lossless compression in X-ray applications reduces image size without impacting quality. This means more images can be taken in a shorter time, resulting in shorter exam times and faster diagnoses for patients.

Real-time fluoroscopy relies on continuous X-rays for applications including cardiology, radiography and neurovascular imaging. The bandwidth demands of transmitting real-time video from the flat panel detector to processing makes it a good candidate for compression techniques.

Similarly, dental imaging is adopting compression techniques as it evolves from still X-ray to more complex 3D applications. Panoramic radiography, or panoramic X-ray, uses small doses of ionizing radiation to capture the entire mouth in a single image, including teeth, jaw structure, and surrounding tissues. Unlike intraoral X-ray where the film is placed inside the mouth, a panoramic X-ray machine consists of an X-ray tube mounted on one side and an FPD on the opposite side. During the examination, the X-ray tube rotates in a semicircle and projects a beam through the patient onto the FPD.

The FPD integrates embedded hardware supporting 1 Gbps image transmission – along with metadata, power, and control data – to processing over low-cost Gigabit Ethernet. As imaging performance demands increase, maintaining the cost of the network infrastructure plays a key role in total system cost advantages for solution providers.

To boost the capabilities of the existing Ethernet infrastructure, the mathematically lossless compression hardware acceleration approach transmits up to 1.5 Gbps throughput rates over existing 1 Gbps infrastructure. The compression technique ensures original and post-compression data are identical while supporting ultra-low latency performance for real-time dental imaging.

Emerging Trends and Compression

There are several emerging application areas that will benefit from efficient compression techniques focused on preserving quality for end-processing. Multi-sensor applications that fuse imaging and data streams require efficient compression to maintain performance over limited infrastructure. The growing deployment of 3D across a range of end-applications, from logistics to robotics and inspection, will also increase bandwidth demands for infrastructure.

Designers have several choices in their approach to compression, and the right choice considers the processing requirements of the application and the cost and performance expectations for end-users.