Hybrid Adaptive Lossless Image Compression Based on Discrete Wavelet Transform

• Published in: Entropy (Basel, Switzerland) Vol. 22; no. 7; p. 751
• Main Author: Starosolski, Roman
• Format: Journal Article
• Language: English
• Published: Switzerland MDPI AG 09.07.2020; MDPI
• Subjects: Algorithms; Compression ratio; Discrete Wavelet Transform; Entropy; entropy estimation; hybrid transform; Image compression; JPEG encoders-decoders; Lossless equipment; lossless image compression; Noise reduction; predictive coding; transform coding; Wavelet transforms; JPEG 2000; entropy estimation; predictive coding; screen content coding; step skipping; discrete wavelet transform; transform coding; lossless image compression; hybrid transform; reversible denoising and lifting step
• ISSN: 1099-4300; 1099-4300
• DOI: 10.3390/e22070751

A new hybrid transform for lossless image compression exploiting a discrete wavelet transform (DWT) and prediction is the main new contribution of this paper. Simple prediction is generally considered ineffective in conjunction with DWT but we applied it to subbands of DWT modified using reversible denoising and lifting steps (RDLSs) with step skipping. The new transform was constructed in an image-adaptive way using heuristics and entropy estimation. For a large and diverse test set consisting of 499 photographic and 247 non-photographic (screen content) images, we found that RDLS with step skipping allowed effectively combining DWT with prediction. Using prediction, we nearly doubled the JPEG 2000 compression ratio improvements that could be obtained using RDLS with step skipping. Because for some images it might be better to apply prediction instead of DWT, we proposed compression schemes with various tradeoffs, which are practical contributions of this study. Compared with unmodified JPEG 2000, one scheme improved the compression ratios of photographic and non-photographic images, on average, by 1.2% and 30.9%, respectively, at the cost of increasing the compression time by 2% and introducing only minimal modifications to JPEG 2000. Greater ratio improvements, exceeding 2% and 32%, respectively, are attainable at a greater cost.