THE AESTHETIC APPEAL OF DEPTH OF FIELD IN PHOTOGRAPHS Tingting Zhang 1 , Harold T. Nefs 1 , Judith Redi 1 , Ingrid Heynderickx 2 1 Delft University of Technology 2 Eindhoven University of Technology ABSTRACT We report here how depth of field (DOF) affects the aesthetic appeal of photographs for different content categories. 339 photographs spanning eight categories were selected from Flickr, Google+, and personal collections. First, we classified the 339 photographs into three levels of depth of field: small, medium, and large. Then, we asked participants to rate the aesthetic appeal of these photographs in random order. We found that aesthetic appeal is only influenced significantly by the content category and by depth of field for animal and sport related photographs. Therefore, we conclude that depth of field should not be regarded as a common criterion for judging aesthetic appeal in different semantic content categories. Index Terms — Depth of field, blur, aesthetic appeal 1. INTRODUCTION We are living in a world where the rapidly growing use of digital cameras and phone cameras, combined with social media such as Facebook 1 , Twitter 2 , and Google+ 3 results in billions of photographs being generated and shared online. Web-based image repositories such as Flickr 4 and Photo.net 5 where professional and amateur photographers can share their work and learn from others’ work are also becoming more and more popular. Managing such a large amount of image material is challenging. Storing, retrieving, and presenting images such that the user quality of experience [1] is high entails determining criteria according to which photographs are preferred by users. Aesthetic appeal of photographs [2-5] has been recently studied in this sense, for various applications such as recommendation systems, retrieval, album management, and also image quality assessment [6-8]. Predicting in an automated way the aesthetic appeal of a photograph implies determining image attributes that influence the visual quality, mapping those attributes into objectively measurable image properties, and finally fuse the latter into an estimation of the image aesthetic appeal [2, 1 www.facebook.com 2 www.twitter.com 3 www.plus.google.com 4 www.flickr.com 5 www.photo.net 9-11]. It is known that evaluation of aesthetic appeal is subjective, highly dependent on the semantic content, on the genre of the images, and also on the viewers’ interest [12, 13], which are difficult to quantify. As a result, many of the existing models which predict aesthetic appeal are based on measures of image properties that can to some extent replicate rules of thumb of good photography. Interestingly, many of these rules have not been empirically proven to actually matter for the aesthetic appeal of photographs. In this paper, we aim to build a bridge between human perception of aesthetic appeal and its automatic evaluation by conducting a study on the importance for aesthetic appeal of such photography elements. We investigated the effect of a typical photographical parameter, namely, DOF, on the aesthetic appeal of photographs. We used a large dataset of consumer photographs spanning eight content categories; the photographs being chosen from web-based image repositories (e.g., Flickr). We performed our investigation in two steps: first, participants were requested to determine whether the photographs had “Small”, “Medium” or “Large” DOF. Then, another group of participants rated all the photographs on aesthetic appeal. Based on these two subjective evaluations, we were able to quantify the impact of the DOF size on aesthetic appeal, depending on the semantic content category of the image. 2. RELATED WORK Since Berlyne’s [14] initial empirical study on aesthetics, many researchers have focused on better understanding human perception of image aesthetics and making the assessment of automatic systems closer to subjective judgments [15]. Cerosaletti & Loui [5] found that in consumer images, semantics of the image subject (e.g. “images with people” vs. “images without people”) as well as attributes of the image such as “main object size”, and “perspective cues” all influence perceived aesthetic appeal significantly. Alongside those, colour, composition, sharpness, perspective cues, form, and spatial organization were also found to be important attributes for the aesthetics of images [6, 16]. Researchers and photographers claim that, apart from the factors mentioned above, DOF also influences the aesthetic appeal of photographs [2, 17, 18]. People working on photography and cinematography use low DOF to enhance an image, such that objects that are out of focus are blurred to a level depending on their distance to the focal plane: an 2014 Sixth International Workshop on Quality of Multimedia Experience (QoMEX) 978-1-4799-6536-6/14/$31.00 ©2014 IEEE 81