The type of media shared through social media channels has shifted from text content to include an increasingly large number of images. Visual traces resulting from people's online social behavior have the potential to reveal insights about our habits, activities and preferences. The role of social network-related factors have been well studied in previous research. Yet, few studies have sought to understand how user behavior in social networks is dependent on the image itself. The goal of my dissertation is to understand how people engage with image content, and I seek to uncover the role of presentation and image content on people's preferences. To achieve this goal, I study the image sharing communities, Flickr, Instagram and Pinterest, using quantitative and qualitative methods. First, I show how colors -- a fundamental property of an image -- could impact the virality of an image on Pinterest. I consider three dimensions of color: hue, saturation and brightness and evaluate their role in the diffusion of the image on Pinterest, while controlling for social network reach and activity. Next, I shift the focus from abstract colors to a higher-level presentation of images. I study the role of filters on the Flickr mobile application as proxies to visual computation. To understand how people use filters, I conduct an interview study with 15 Flickr mobile users about their filter use. I analyze Flickr mobile images to discover the role of filters in engaging users. Presentation is not the only factor that makes an image interesting. To gain deeper insights in what makes an image more engaging in social image sharing sites, I study the images of people on the Instagram network. I compare images of people with those that do not have faces and find that images with human faces are more engaging. I also look at the role of age and gender of people in the image in engaging users. Finally, I examine different content categories, with and without filters, and study the impact of content category on engagement. I use large-scale data from Flickr and interviews with Flickr mobile users to draw insights into filter use and content engagement. This dissertation takes a first step toward understanding content and presentation of images and how they impact one aspect of user behavior online. It provides several theoretical and design implications for effective design, creation and imposition of rules on image sharing communities. This dissertation opens up a new direction for future research in multimedia-mediated communication.