Search

Session 3B-Blanes-CCSDS 123.0-B-2 compression standard cross-validation and fuzzing

In 2019, the 123.0-B-2 standard titled “Low-Complexity Lossless and Near-Lossless Multispectral and Hyperspectral Image Compression” was presented. It introduced a near-lossless operation mode and a new state-of-the-art entropy encoder to provide efficient coding of the low-entropy data often produced by the new in-loop quantizer included in the standard. This article shows that tunning properly some parameters during the coding process, the weight update stage of a CCSDS 123.0-B-2 predictor can be bypassed, yielding a faster encoder at expenses of likely compression ratios. Experimental results show the coding performance penalization bypassing the weight updating, compared with a regular coding process --when the weights are updated-, for lossless and near-lossless compression. Coding performance analysis is provided for Hyperion, Aviris, Modis and Airs instruments. Results suggest that, bypass the weight updating produces a lossless coding performance penalization about 4\%. For near-lossless, a higher coding penalty is produced as the step size, in the inner loop quantizer of the predictor, is larger.