Elsevier

Appetite

Volume 96, 1 January 2016, Pages 166-173
Appetite

Standardized food images: A photographing protocol and image database

https://doi.org/10.1016/j.appet.2015.08.041Get rights and content

Highlights

  • We developed a standardized food image photographing protocol.

  • Liking, perceived calories and healthiness were scored by adults and children.

  • High and low calorie images differed in perceived calorie content and healthiness.

  • The food images are suitable for use across countries.

  • The protocol and standardized food images are available at nutritionalneuroscience.eu.

Abstract

The regulation of food intake has gained much research interest because of the current obesity epidemic. For research purposes, food images are a good and convenient alternative for real food because many dietary decisions are made based on the sight of foods. Food pictures are assumed to elicit anticipatory responses similar to real foods because of learned associations between visual food characteristics and post-ingestive consequences. In contemporary food science, a wide variety of images are used which introduces between-study variability and hampers comparison and meta-analysis of results. Therefore, we created an easy-to-use photographing protocol which enables researchers to generate high resolution food images appropriate for their study objective and population. In addition, we provide a high quality standardized picture set which was characterized in seven European countries. With the use of this photographing protocol a large number of food images were created. Of these images, 80 were selected based on their recognizability in Scotland, Greece and The Netherlands. We collected image characteristics such as liking, perceived calories and/or perceived healthiness ratings from 449 adults and 191 children. The majority of the foods were recognized and liked at all sites. The differences in liking ratings, perceived calories and perceived healthiness between sites were minimal. Furthermore, perceived caloric content and healthiness ratings correlated strongly (r ≥ 0.8) with actual caloric content in both adults and children. The photographing protocol as well as the images and the data are freely available for research use on http://nutritionalneuroscience.eu/. By providing the research community with standardized images and the tools to create their own, comparability between studies will be improved and a head-start is made for a world-wide standardized food image database.

Introduction

The regulation of food intake has gained much research interest because of the current obesity epidemic. Accordingly, a broad range of studies from several fields of research has investigated the determinants of (un)healthy eating behavior. This includes studies in psychiatry (Cowdrey, Finlayson, & Park, 2013), psychology (Junghans et al., 2013, Veling et al., 2013), food-related neuroscience studies (Kringelbach and Rolls, 2004, Simmons et al., 2005, Tiggemann and Kemps, 2005, Van Der et al., 2014; L. N. van der Laan et al., 2011, van Meer et al., 2014) consumer research (Laura N van der Laan, Hooge, de Ridder, Viergever, & Smeets, 2015), nutritional sciences and sensory sciences. Many of these studies use food images as a substitute for real foods to investigate the determinants of eating behavior caused by visual food cue exposure. Visual food cues are omnipresent in our environment and many of our dietary decisions are based on the sight of food due to learned associations between visual food characteristics and their post-ingestive effects (Lappalainen and Sjoden, 1992, Wardle, 1990). Therefore, food images are a good and easy-to-use alternative to real food to study responses to visual food exposure.

A wide variety of food images has been used. For example, in neuroimaging studies alone, images vary from single food items to filled plates or even complete pots with food (Brooks et al., 2012, Frank et al., 2010, Führer et al., 2008, Malik et al., 2008, Mehta et al., 2012, Siep et al., 2009, Simmons et al., 2005). Examples of food stimuli are shown in Fig. 1. Some studies used images made by the researchers themselves, often poor in lighting; other studies used images from commercial stock photograph websites (Killgore et al., 2003, Killgore and Yurgelun-Todd, 2005, Rothemund et al., 2007, Siep et al., 2009), which are of high photographic quality but highly diverse in portion size and presentation. Commonly, studies only report a brief description of the images used, which complicates replication (Beaver et al., 2006, Cornier et al., 2007, Davids et al., 2009, Holsen et al., 2006, Holsen et al., 2005, LaBar et al., 2001, Miller et al., 2007, Santel et al., 2006, Schienle et al., 2009). In an attempt to stimulate replication studies, a few groups have started to share their image sets online (e.g. (Blechert, Meule, Busch, & Ohla, 2014)). These images are often collected from the internet, pasted on a plain background (white, gray or black) and adjusted in brightness and contrast. The downsides of this are the lack of shadows (which gives the food the appearance of hanging in the air), the different angles at which foods are depicted, different magnifications (e.g., a sandwich shown the same size as a tomato), the lack of a visual reference item such as a plate, and image deterioration owing to the adjustments in brightness and contrasts, which can affect the attractiveness of the food items (e.g. (Knebel, Toepel, Hudry, le Coutre, & Murray, 2008)). Until now, only a few image sets have been used in multiple studies, usually limited to use in the same research lab (e.g. (Holsen et al., 2006, Holsen et al., 2005, LaBar et al., 2001) and (W. D. Killgore et al., 2003; William DS Killgore and Yurgelun-Todd, 2005, Rothemund et al., 2007)). In conclusion, the images used in food research are highly variable and only seldom re-used. This is unfortunate, because it introduces within- and between-study variability and hampers comparison and meta-analysis of results (L. N. van der Laan et al., 2011). Thus, there is a clear need for a high quality food image set, standardized in luminance and presentation, which researchers can use for multiple research purposes.

For across-nation comparison of results, an important point to consider is that each country has its own food specialties and availability of fruits and vegetables. This complicates the (re)use of one set of images in experiments conducted in different countries. Because results are often compared between countries and the number of international multicenter studies is increasing, there is a need for a standardized food image set suitable for use across nations. Furthermore, with the increasing interest in age differences in eating behavior, it is important that images are suitable for research in children as well as adults.

The most important characteristic of food images is the recognizability of the portrayed food. Additional image characteristics that can be of interest are for example liking (e.g. to ensure equal attractiveness across food categories), the perceived and actual caloric content, and perceived healthiness (e.g. especially important when a high versus low calorie paradigm is used). Besides these characteristics, the ability of an image to exert a similar response as real food (e.g. do liking ratings of the images increase with hunger) is also of importance.

To address the need for high quality standardized food images we here present our easy to use standardized photographing protocol which enables researchers with a broad range of research objectives to generate additional study- or country-specific high resolution food images. In addition, we used the protocol to create images for a standardized food viewing fMRI task that is being used in several European countries. For these images, data was collected from adults and children to facilitate image selection on both visual appearance and food characteristics (e.g. caloric content, recognizability ratings, liking ratings etc.). We thus make important first steps in image standardization and validation across countries.

Section snippets

Standardized photographing protocol

The images were created with the use of a high resolution digital single lens reflex camera which was mounted on a tripod. The focal length used was 32.0 mm. In order to depict each food as attractive as possible, shutter speed and aperture were automatically adjusted for each food. The pictures were taken in a 60 × 60 × 60 cm photo studio (i.e., a cubic photo tent made from snow white fabric which softens and reflects light). The tent was completely closed. An incision was made in the front to

Standardized food images

Fig. 4 shows examples of the images created on the different sites. Although different types of food were photographed, this figure clearly shows the similarity of the images produced at different sites that can be achieved.

Food image recognizability

As previously mentioned, the 80 images presented in this study were selected on recognizability of adults in Scotland, The Netherlands and Greece (see Section 2.2). The mean ± SD percentages of recognized ratings per image for these sites were: SCT = 98.9 ± 2.4;

Discussion and conclusion

We made a photographing protocol suitable for creating high-quality standardized food images. Using this protocol we created a food image set and tested a subset of 80 images in several European countries in both adults and children. Currently, the total set comprises 370 images (including country specific images and different preparations) of 260 different food items and the number of images is still growing.

A large proportion of the images was recognized by the sites not included in the image

Acknowledgments

This work was financially supported by the European Union Seventh Framework Programme (FP7/2007–2013) for research, technological development and demonstration under grant agreements 266408 (Full4Health, www.full4health.eu), 266044 (I.Family, www.ifamilystudy.eu) and 607310 (Nudge-It, www.nudge-it.eu).

We thank our colleagues from the University of Aberdeen, University of Leeds, Harokopio University of Athens, the Leibniz-Institute for Prevention Research and Epidemiology (BIPS)/Bremen

References (37)

Cited by (113)

View all citing articles on Scopus
View full text