Many people (especially women) tend to take a long time for shopping. From the results of our preliminary questionnaire, we conclude that many female students look many products to find the best product, while enjoying the shopping processes themselves. This paper presents a user interface for exploratory apparel product search inspired by the questionnaire results. Our study supposes that each product is assigned with a pre-defined keyword, and prepares icons associated to the combination of the keywords. The interface displays various icons in a display space to show the diversity of the products. The icons are generated in the preprocessing. They are designed specific to this study, as the overlay of images representing shapes, colors, and texture patterns of the products. When a user selects an interested icon, the interface switches the display to a set of images corresponding to the selected icon, such that the user can visually compare the similar products. It acts real shopping behavior because we often firstly look over the shops to understand the diversity of products, and then close up the particular groups of the products. The interface also estimates the preferences of users from their selection of icons, and applies an evolutionary computation algorithm which adjusts the selection of icons to their preferences. This paper introduces a user experiment to demonstrate the effectiveness of the presented interface, and discusses how the interface adopts to (especially women's) shopping psychology.

View full abstract

Portrait icons have been increasingly used in various scenes including social networking services (SNS). Popular commercial systems to electronically create cartoons require users to select "part illustration images" similarly looking to the users' faces. These systems provide sample images to be selected by the users, and synthesize the selected images to complete the cartoon generation. However, this approach is often difficult to create cartoons similarly-looking to the input real face images when there are no sample images which appropriately represent the features of input faces. In this paper, we propose the technique which generates cartoons applying a morphing method to multiple sample part illustration images. As a result, this technique relieves the dependency of quality on the adequacy of sample part illustration images. Also, the technique realizes to generate part illustration images that well reflect characteristics of the input real face images.

View full abstract

Information visualization is an effective approach to analyze time-varying data in our daily lives. We commonly represent time-varying values applying polyline charts or heatmaps; however, it is difficult to simultaneously observe short-term features of time-varying values and cluster transitions while applying either polyline charts or heatmaps. This paper proposes a storyline-based visualization technique for consecutive numerical time-varying data. Storyline is a visualization technique to show associative features among elements over time. Our technique first measures similarity of elements in each time-step, and divides the elements into clusters. The technique then defines the cluster layout by matching corresponding clusters between two adjacent time-steps, and draws similar elements as proximity storyline. Reflecting transparency on storyline as a visual variable, the technique also emphasizes the amount of line changes. Moreover, the technique provides a user interface so that users can interactively select interesting parts on storyline, and explore the numerical values by observing a polyline-based visualization. We believe it is important to focus on elements which switch their clusters frequently. We suppose that by making the appearances of numerical changes prominent based on the amount of changes, a user would be able to effortlessly pay his/her attention to where those changes occurred. This easy recognition of numerical changes would lead to further focused investigation on the causality through examination of the original numerical values and other associated information.

View full abstract

This paper proposes a new method to enable a user to interactively edit an existing object displayed in a virtual scene on a free-viewpoint video in real time simultaneously while capturing its original real-world scene by a camera array. The proposed method consists of the pure extraction method and the editing method. The pure extraction method is based on synthetic aperture imaging and purely extracts the whole shape of a target object that a user wants to edit on a free-viewpoint video even if the object is partially occluded by other objects. The editing method enables the user to interactively edit the extracted target object by 2D geometric transformations directly on the video in which the object is displayed in front of its background. The background is always displayed as it is in the real-world scene with no influence by the editing in the virtual scene; for example, any empty regions are not caused in the background in the virtual scene by the editing even if the background contains moving objects. Some experimental results show a good ability of the proposed method.

View full abstract

Image matting is a technique used to extract a foreground object from an input color image by estimating its opacity mask, which is called an alpha matte. Many previous methods have estimated a poor-quality alpha matte when a foreground object and its background have similar colors. To overcome this problem, we treat an image matting problem as a binary classification problem to classify unknown pixels into foreground and background pixels. In this study, we propose a learning-based matting method based on binary classification. In our method, a binary alpha matte is first obtained by a binary classifier. The binary alpha matte is then iteratively refined to a high-quality alpha matte by repeat use of the classifier. Although we used a support vector machine classifier and the closed-form matting method, an important merit of our method is that it provides a general mechanism to refine an alpha matte by combining various binary classifiers and matting methods. The excellent performance of our method was revealed in several experiments, especially, for input images having similar foreground and background colors as well as objects with small holes.

View full abstract

In this paper, we propose and implement a new method to create generative art using soot of candle that is controlled by a computer. This system is called 'Sootoid'. A Sootoid uses two motions for drawing. One is a movement of a candle that is based on a program, and it has repeatability. The other one is a movement of a flame of a candle that is based on nature phenomena, and it does not have repeatability. Therefore, while the most of generative arts use a random number to make a noise of system; as against Sootoid uses a randomness of natural phenomena of a system to generate a noise.

View full abstract

The making of the cast list is an intermediate step in the video production. It is a list of the stages and the characters appearing in each scene. In Japan the cast list is derived from the screenplay manually, with someone getting out the information from the screenplay. To help with this step, the system studied in this paper proposes a structured screenplay which makes it possible to extract the necessary information automatically. With a structured screenplay, the writer makes annotations while analyzing the content. However, the annotations added at this step still do not provide enough information to produce a cast list. Therefore, we propose a system that is able to extract the additional required information. We have conducted experiments with producing a cast list from a structured screenplay. With the results of these experiments, we can confirm that it is possible to extract the information necessary to produce a cast list from a structured screenplay. We were also able to confirm that, by using this system, the time used for getting out the necessary information compared to working without this system, is shortened by about 30-40 percent. It has become evident that by using a structured screenplay together with this system, the production of cast lists becomes more efficient. Therefore this system will be helpful in production of video content.

View full abstract

Balloon modeling is the artistic technique of twisting balloons into various shapes. We have proposed a method for formalization and rating difficulty of balloon modeling made of one balloon based on graph theory, and developed a system for supporting balloon modeling making based on the method. In this paper, we propose a method for separating a graph which is defined as a structure of balloon modeling made of multiple balloons into the minimum number that is needed for making. We also develop a design support system that users can set shape and color of balloons based on proposed method.

View full abstract