Study of naturalness in tone-mapped images

Abstract

Nowadays, images can be obtained in various ways such as capturing photos in single-exposure mode, applying Multiple Exposure Fusion algorithms to generate an image from multiple shoots of the same scene, mapping High Dynamic Range images to Standard Dynamic Range (SDR) images, converting raw formats to displayable formats, or applying post-processing techniques to enhance image quality, aesthetic quality,.. . When looking at some photos, one might have a feeling of unnaturalness. This paper deals with the problem of developing a model firstly to estimate if an image looks natural or not to humans and the second purpose is to try to understand how the unnaturalness feeling is induced by a photo: Are there specific unnaturalness clues or is unnaturalness a general feeling when looking at a photo? The study focuses on SDR images, especially on tonemapped images. The first contribution of the paper is the setting of an experiment gathering human naturalness opinions on 1,900 SDR images mainly obtained from tone mapping operators. Based on the collected data, the second contribution of the paper is to study the efficiency of different feature types including handcrafted features and learned features for image naturalness analysis. A binary classification model is then developed based on the determined features to classify if an image looks natural or unnatural.

Computational Analysis of Correlations between Image Aesthetic and Image Naturalness in the Relation with Image Quality

“…In some studies, IN features have been employed for IQA [ 17 , 18 , 19 ]. Moreover, in [ 1 ], IN is based on artifacts induced by some image processing methods (such as halos, blur, lost details) and on the individual feeling (memory, opinion, background).…”

“…In this research, an IA dataset coming from [ 35 ] and an IN dataset coming from [ 1 ] are considered. On the one hand, the IA dataset contains 1200 high aesthetic images and 1200 low aesthetic images coming from the CUHKPQ dataset [ 16 ].…”

“…On the side of IN, this notion is both related to artifacts induced by some image processing algorithms and to an individual feeling [ 1 ] (see Figure 1 (6) and Figure 3 b). Regarding artifacts, IN is affected by strong visible clues detected by viewers’ eyes, so the unnaturalness feeling comes from annoying artifacts induced by camera sensors, image processing algorithms (compressing, tone-mapping), image format or file transfer (see bottom right photo in Figure 3 b).…”

Image Aesthetic Assessment Based on Image Classification and Region Segmentation

“…The local features are computed from ROIs, background and regions split by symmetry rules, landscape rule, rule of thirds (see Figure 5 ). At the next step, the feature reduction algorithm introduced in [ 36 , 40 ] is applied on 1200 large field images and 1200 close-up images coming from the CUHKPQ dataset [ 25 ] in which a half of them is used in the training phase ( ) and the remaining is used in the testing phase ( ). After the most relevant features are selected, those features are analyzed to remove overlapping features and to optimize the feature set.…”

“…The feature reduction algorithm described in [ 36 , 40 ] is run on 1200 large field images and 1200 close-up images coming from the CUHKPQ dataset to select the 925 most relevant features (the highest classification performance is obtained with those features) among the 4096 features learned by the VGG16.…”