Importantly, we then apply a weight step to divide the high-frequency spectral coefficients by the overall luminance (DC component) to allow for more aggressive compression of less important data. Rather than inventing a completely new file type that is then fed into the codec, this method uses a compression engine and standardized functionality JPEG XL An image format that stores specially prepared spectral data.
Makes spectral images easier to manipulate
Researchers say the large file size of spectral images is a real barrier to industry adoption that benefits from its accuracy. Smaller files mean lower transfer times, lower storage costs, and more interactive operation of these images without specialized hardware.
The results reported by the researchers appear to be impressive. With that technique, the spectral image file is reduced by 10-60 times compared to standard OpenEXR lossless compression, making it comparable to the size of a normal high quality photo. It also retains major OpenEXR features such as metadata and high dynamic range support.
Some information is sacrificed in the compression process, but this is “lossed” form, but researchers first design to discard the most prominent details, focusing on compression artifacts on critical high-frequency spectral details to maintain important visual information.
Of course, there are some limitations. We translate these findings into extensive practical use for the continuous development and improvement of software tools that handle JPEG XL encoding and decoding. Like many cutting-edge formats, initial software implementations may require further development to completely unlock all features. It’s work in progress.
Also, Spectral JPEG XL dramatically reduces file size, but its lossy approach can bring down some scientific applications. Some researchers using spectral data may easily accept the trade-offs for the practical benefits of small files and faster processing. Others that handle particularly sensitive measurements may need to look for alternative storage methods.
For now, this new technique is primarily of interest to specialists such as scientific visualization and high-end rendering. However, as the industry from automotive design to medical imaging continues to generate larger spectral data sets, such compression techniques can help make these large files more practical.
