The effect of image content on detail preservation and file size reduction in lossy compression

Abstract

Listen

Translate article

To demonstrate the effect of image content on image detail preservation and file size reduction. The first set, containing 16 in vitro images with variable projection geometry, exposure time, bone level and number of teeth, was compressed with three compression modes: JPEG quality factor (JPQF), JPEG2000 quality factor (J2QF) and JPEG2000 compression ratio (J2CR). Image detail degradation was evaluated by local mean square error (MSE) on a standardized region of interest (ROI), containing bone. The second set, containing 105 clinical bitewings, was compressed with the same compression modes at 3 quality factors/compression ratios and local MSEs were calculated on two ROIs, containing bone and crown. For the first image set, nearly constant MSE was found for the JPQF and J2QF compression modes, while file size depended on projection geometry, exposure time, bone level and the number of teeth. In contrast, file size reduction was nearly constant for the J2CR compression mode, while MSE depended on the abovementioned factors. Similarly, for the second image set, nearly constant MSE and variation of file size reduction were found for JPQF and J2QF but not for the J2CR compression mode. All of these results were consistent for all three quality factors/compression ratios. Constant image detail preservation, crucial for diagnostic accuracy in radiology, can only be assured in QF compression mode in which the file size of the compressed image depends on the original image content. CR compression mode assures constant file size reduction, but image detail preservation depends on image content.