JAI Xpress lossless compression for Gigabit Ethernet

As today’s CMOS cameras continue to increase in resolution and speed, JAI have introduced a lossless compression feature in order to use the sensor at its maximum capacity while remaining GigE without having to shift to 5GigE, 10GigE or 25GigE. This allows higher performance without extra cost and without the complication that these faster interfaces introduce. JAI Xpress allows users to keep the benefits of the GigE Vision standard but squeeze every ounce of performance from the camera.

What is Xpress?
Xpress provides a lossless compression algorithm that reduces the size of image data without losing any information and so allowing a full reconstruction later on the host PC with exactly the same quality as captured. Compression is performed in the camera’s FPGA using the principle of image redundancy. Specifically, the Xpress algorithm deconstructs images into small blocks of pixels and finds other blocks in the image with the same pixel pattern. By representing redundant blocks with short codes, the total size of the “encoded” image data can be significantly reduced before it is output from the camera. The data that encapsulates the image is smaller and so frames are transmitted at a higher rate using standard GigE Vision bandwidth.

How much faster depends on the amount of redundancy in the image, ranging from small increases for extremely “busy” images, to speed increases of 100% or more in images with a high degree of homogeneity. For example, decreasing the size of an image file by 30% results in a frame rate increase of approximately 42% over a standard GigE Vision connection. Alternatively, the smaller file sizes possible with Xpress can enable you to increase the number of cameras sharing the same network, while maintaining your current frame rates.

In order to compress GigE Vision data, users can include a few functions supplied in JAI's SDK into their application code. The Xpress compression software is available for downloading from the JAI website.