Flame is one of the most frequently used natural phenomena in the field of computer graphics (CG). However, due to its complexity, a simple model without considering combustion process has been frequently used. A reaction with oxygen, which is one of important phenomenon for flame, has not been considered in the previous researches, while a flame in real-world would be intensed by supplying a oxygen. Also, the unique fluctuation of the flame such as the temperature fluctuation due to the incomplete combustion has not realized. In this research, we model the concept of reaction with oxygen and the chemical phenomena that occur in the combustion process, and propose a new flame simulation method using it. The flame and oxygen are simulated independently by both the particle method and the lattice method, and the physical quantity is exchanged between the two simulations to handle the combustion considering the chemical reaction. We realized the fast simulation by using a GPU parallel computation. As a result, we confirmed that the simulation of a flame with combustion process including the oxygen can be performed at 10～20 fps.

Number of endoscopic surgery trends upward because the postoperative recovery is fast and the patient can be discharged immediately. Although endoscopic surgery is less burden on patients than open surgery, its difficulty is high for unskilled doctors. Recently, an 8K resolution endoscope camera has been developed to improve the technical level of endoscopic surgery. An 8K endoscope camera is expected to reduce the difficulty of surgery because it can obtain a lot of information about texture and depth of the organ. However, an 8K endoscope camera has a shallower depth of field than traditional cameras, and it is difficult to simultaneously focus on multiple objects such as blood vessels and sutures. In this paper, the method which makes 8K endoscopic videos with insufficient focusing sharp is proposed. And then, in order to quantify the performance of the proposed method, the visibility of sutures used in surgery and the image quality degradation due to noise were evaluated by subjective assessments. As a result of the experiments, it is shown that image quality degradation was avoided, and that sutures could be found in a shorter time.

An ambigram is a word which can be also read from a different direction. Generation of ambigrams, however, requires not only enormous work but also design skills for maintaining their readability. This paper, therefore, aims at ambigram generation for a pair of Hiragana. First, characters are thinned, and strokes are extracted and simplified to obtain feature points of the characters. Next, subgraphs are identified from the graphs with the feature points as nodes, common structures between the characters are extracted by verifying isomorphism of the subgraphs, and ambigram structure is generated by affine transformation based on the correspondence between common structures. Finally, stroke thicknesses are exchanged with brush style strokes to improve their readability and 1,081 ambigrams were generated from pairs of 46 hiragana characters. Furthermore, readability of ambigrams was evaluated. In case of a single ambigram, 325 out of 724 ambigrams which can be given by hand crafted were readable. Moreover, 93 out of 357 ambigrams which are not pointed out to be generated were readable. Words composed of 4 hiragana characters including one ambigram with a low readability also revealed that 216 out of 246 words were readable.

Kimekomi is a traditional craft produced by attaching cloth to a wooden support structure with engraved grooves. Using the production kit sold in handicraft shops, even beginners can easily create one. However, it is difficult for beginners to design an original one. In this paper, we propose a design support system for Kimekomi for beginners. We set plane and cube as our target. A user designs a Kimekomi appearance using our system, and the system automatically generates the support structure of the Kimekomi product based on the design. The system also visualizes the final result. The system estimates production time and present it to the user. The user can design an original Kimekomi product appropriate for the user’s skill level.

In this study, we construct a deep learning model that generates realistic images and animations of plants from simple point inputs that specify the contents of images. In conventional image generation by deep learning, a rough input may be difficult because the input and output images need to correspond one-to-one for each pixel. In addition, a large amount of input data is required to generate an animation. On the other hand, by extracting and manipulating attributes of the image for continuously changing an image, it is difficult to obtain a high-quality plant animation in which details are clearly expressed. In this study, we construct a two-stage deep learning model using point labels as input. As a result, high-quality images and animations that plants smoothly change can be generated from a small amount of learning data. Our quantitative evaluation showed that high-quality images were obtained that were clearer and less biased in appearance attributes such as leaf arrangement and the size of the plant.