How Good is Too Good? A Subjective Study on Over Enhancement of Images

Abstract

For a long time different studies have focused on introducing new image enhancement techniques. While these techniques show a good performance and are able to increase the quality of images, little attention has been paid to how and when overenhancement occurs in the image. This could possibly be linked to the fact that current image quality metrics are not able to accurately evaluate the quality of enhanced images. In this study we introduce the Subjective Enhanced Image Dataset (SEID) in which 15 observers are asked to enhance the quality of 30 reference images which are shown to them once at a low and another time at a high contrast. Observers were instructed to enhance the quality of the images to the point that any more enhancement will result in a drop in the image quality. Results show that there is an agreement between observers on when over-enhancement occurs and this point is closely similar no matter if the high contrast or the low contrast image is enhanced.

Preventing Over-Enhancement Using Modified ICSO Algorithm

“…We compared the performance of the proposed approach with nine other state-ofthe-art methods namely, Adaptive Gamma Correction with Weighting Distribution (AGCWD) [35], Contrast Limited Adaptive Histogram Equalization (CLAHE) [36], Dynamic Piecewise Linear Transformation (DPLT) [37], Exposure based Sub-Image Histogram Equalization (ESIHE) [10], Low-Complexity Algorithm for Contrast Enhancement (LCACE) [38], probabilistic method for image enhancement [39], Automatic and Parameter-Free Piecewise Linear Transformation (APFPLT) [40], image enhancement with Semi-Decoupled Decomposition (SDD) [41], and swift algorithm [42] morphology, and HVS-inspired. In our experiment 30 original images from the SEID dataset [32] were selected. The contrast level of these images were then increased and decreased significantly resulting in 60 different (30 low-and 30 high-contrast) images.…”

“…Although a wide range of image enhancement methods focused on image contrast have been introduced, they are mainly focused on low contrast images or proposed to improve the quality of one type of degraded image [32]. In this study, we aim to introduce a new method to enhance the quality of both low-and high-contrast images.…”

“…The Distribution of MOS values for each enhancement technique applied on the high-(a) and low-contrast and brightness (b) images. Image labels in the plot correspond to image names in the SEID dataset[32]. Sample images from the dataset are also shown in the plot.…”

Individual Contrast Preferences in Natural Images

“…Cherepkova et al [15] show that observers have distinguishable patterns with regard to contrast distortion when judging image quality, where some observers rate images with higher contrast to have higher quality and vice versa. Contrast is a fundamental image attribute that can influence how the image is perceived by observers and has a significant impact on IQA [16] and enhancement [17]. The effect of contrast on perceived image quality was discovered in early works; an example of such a subjective experiment was reported in the work of Roufs and Goossens in 1988 [13] and by Roufs et al in 1994 [18].…”

Analyzing the Variability of Subjective Image Quality Ratings for Different Distortions

“…While IQMs show a high correlation with MOS [2], there still exists a high variability between the scores given by different observers to the same image [3].This issue demonstrates the need of an IQM, which is able to work for each individual observer. In spite of such a clear need, due to its complexity, a personalized IQM has rarely been investigated in the research community [4].…”