Towards measuring the visualness of a concept

Abstract

In this paper, we propose a new method to measure the visualness of a concept. The visualness of a concept is generally defined as what extent a concept has visual characteristics. Even though the visualness of a concept is important and useful for various image search tasks, it has not received much spotlight yet. In this work, we especially focus on how to measure the visualness of a complex concept such as "round table", "dry bed" rather than a simple concept like "ball", "apple". To measure the visualness, we first collect sample images of a complex concept using web image search engines, and then group the images based on the visual features. Finally, we compute visual purity and weighted entropy of the clusters, which will act as a visualness score for the concept. Through various experiments, we show and discuss interesting results about the visualness of a concept.

Defining Visually Descriptive Language

“…(ii) can these 'visually descriptive' text segments be identified automatically within documents which may consist of predominantly 'non-visual' text? To be able to answer these questions, we first require a robust, inter-subjectively reliable definition of 'visually descriptive' text. Although previous work exists that models the 'visualness' of terms or concepts from images (Yanai and Barnard, 2005;Jeong et al, 2012), they are presented without an explicit definition apart from the intuitive notion that a visual term should exhibit some consistent visual characteristics across different objects. To our knowledge, the only work that explicitly proposes a definition for visually descriptive text is that of Dodge et al (2012), where noun phrases within an image caption are classified as to whether or not they are depicted in the corresponding image.…”

[Invited Paper] A Review of Web Image Mining

“…Sun et al 118) proposed to quantify tag representativeness of visual content of social images, which is equivalent to visualness. Jeong et al 55) proposed a method to compute visualness based on clustering, which took account of inter-and intra-cluster purity as well as the entropy. Xu et al 136) clustered Web images into small visually coherent clusters, "visualsets" which are similar to Visual Synsets 124) , and then evaluated visualness of each visualset.…”

Biomedical image representation approach using visualness and spatial information in a concept feature space for interactive region-of-interest-based retrieval

“…An entropy of an image patch (e.g., concept) can be valuable to determine its importance of visualness for image representation. 35 Entropy can be considered as the average number of bits one needs to represent a symbol in a stationary system, where the limited source symbols have fixed probabilities of occurrence. 32 In other words, an entropy is a quantity that is used to describe the degree of randomness of an image.…”