A short review of well-known image codecs and observations and root cause of some strange behaviors in image compression codecs

• Published in: Signal, image and video processing image and video processing, 2020-04, Vol.14 (3), p.583-591
• Main Authors: Kwan, Chiman, Larkin, Jude, Budavari, Bence, Tran, Trac D.
• Format: Article
• Language: English
• Subjects: Behavior; Business metrics; Codec; Compression ratio; Computer Imaging; Computer Science; Experiments; Image compression; Image Processing and Computer Vision; Image quality; Multimedia Information Systems; Original Paper; Pattern Recognition and Graphics; Performance measurement; Signal,Image and Speech Processing; Students; Teachers; Vision
• ISSN: 1863-1703; 1863-1711
• DOI: 10.1007/s11760-019-01591-5

In this paper, we present a short review of some well-known image codecs in the literature and summarize a systematic study that determines the root cause of some puzzling observations in image compression experiments. Moreover, we propose a methodology to determine whether an image is genuine or not, meaning that whether or not a given image has been compressed and decompressed before and by which codec. In image compression class projects, students may observe some strange behaviors when they use some images with unknown quality in compression experiments. That is, some performance metrics from a mediocre codec such as JPEG may have exceptionally high values at certain compression ratios as compared to other high performing codecs. This confusing behavior may be overlooked by instructors, and students may never understand why this is happening. We will first highlight this anomalous behavior. We will then use experiments to systematically determine the root cause, which is due to image quality. In other words, if one uses a previously compressed and decompressed image in some compression experiments, it is highly likely that some strange behaviors in the performance metrics will show up. Our findings include the determination of the root cause of a puzzling phenomenon in image compression experiments and some sound advice to instructors, tutors, and students on how one can prevent such behaviors from occurring. We also developed a methodology to determine whether an image is genuine or not.