Image Aesthetic Assessment Based on Image Classification and Region Segmentation

Abstract

The main goal of this paper is to study Image Aesthetic Assessment (IAA) indicating images as high or low aesthetic. The main contributions concern three points. Firstly, following the idea that photos in different categories (human, flower, animal, landscape, …) are taken with different photographic rules, image aesthetic should be evaluated in a different way for each image category. Large field images and close-up images are two typical categories of images with opposite photographic rules so we want to investigate the intuition that prior Large field/Close-up Image Classification (LCIC) might improve the performance of IAA. Secondly, when a viewer looks at a photo, some regions receive more attention than other regions. Those regions are defined as Regions Of Interest (ROI) and it might be worthy to identify those regions before IAA. The question “Is it worthy to extract some ROIs before IAA?” is considered by studying Region Of Interest Extraction (ROIE) before investigating IAA based on each feature set (global image features, ROI features and background features). Based on the answers, a new IAA model is proposed. The last point is about a comparison between the efficiency of handcrafted and learned features for the purpose of IAA.

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

“…A deep learning approach might be a good solution to describe the subjective aspects. Various studies on IA using deep learning have been proposed, such as the Image Aesthetic Assessment (IAA) model based on the combination of a retrieval system and a deep Convolutional Neural Network (CNN) presented in [ 32 ], the double-column deep CNN architecture using two parallel CNNs based on global and local features proposed in [ 33 ], a CNN including 3 kinds of layers: transferred layers, scene convolutional layers and fully connected layers, evaluating the IA of multi-scenes in [ 34 ], the IAA model based on the deep learning technique, image classification and image segmentation introduced in [ 35 ]. Moreover, Hii et al [ 20 ] proposed a deep model exploiting multiple inception blocks pooled by global average pooling layers.…”

“…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 ].…”

“…Besides that, an IAA model based on the IA feature set learned from [ 35 ], and an INA model based on the IN feature set coming from [ 1 ] are considered in this paper because of their good performances. First, the IA feature set contains 1024 global features (features learned from the whole image) learned by a deep CNN having a typical architecture with an input layer, an output layer and five convolutional blocks.…”

“…Similarly, the most correlated IA feature to the j -th IN feature is determined as the IA feature having the highest correlation ( ) to the IN feature as in ( 2 ), in which 1024 is the number of IA features. Histograms with ten bins are built based on the highest correlation of each feature computed in the IA dataset [ 35 ] and the IN dataset [ 1 ], as in Figure 4 and Figure 5 . …”

Multi-Item Assessment of Physiognomic Diversity of Geocomplexes as a Comprehensive Method of Visual-Aesthetic Landscape Assessment

“…A large number of features (attributes) of landscapes and views are evaluated [23], e.g., in the study [24], twenty five indicators were used. Automatic landscape aesthetic evaluation methods are also advanced [25]. In general, indicators related to landscape diversity (biodiversity, geodiversity, etc.…”