We aim to estimate the depth of microscopic objects by image processing, and describe a depth-from-focus (DFF) method that uses multiple images taken with different focus lengths. DFF methods usually estimate the depth at a pixel by evaluating image sharpness among the images, but they are not good at handling texture-less regions and regions including strong edges because image sharpness is not properly obtained in such regions, which results in accuracy deterioration. To address the problems, we focus on a relation between the luminance of a target pixel and neighboring pixels’ mean value, and propose a sharpness measurement method under the assumption that the relation follows Poisson distribution. In addition, our method uses multiple window sizes when taking into account the neighboring pixels to improve accuracy, and integrates evaluated focus measures by using a DFF algorithm of Fommer et al. We show the effectiveness of our method through numerical results using simulation images and quantitative results using actual images.

View full abstract

3D shape analysis has become an important task in computer vision due to the spread of low-cost 3D sensor. In this background, local feature point extraction is one of the most important techniques, it selects points that have repeatability and distinctiveness. The conventional methods mainly search the points on salient surfaces compared with neighborhood. However, the points on salient surfaces have a problem that is low repeatability with the sensor noise or occlusion. Therefore, we noticed the planar surface that has high repeatability. The proposed method estimates a lot of plane equations from a point cloud, then obtains virtual feature points from the combinations of them. The virtual feature point is called Virtual Keypoint Of Polyhedron (VKOP), its feature descriptor calculated from the combinations. The results of the experimental evaluations showed that the proposed virtual feature points outperforms the conventional feature points in about equal processing time.

View full abstract

As an effective reconstructing technique to observe large sized holographic recovered images in the wide viewing area, we have studied a projecting recovering process of holographic images with the use of the spatial screen. Recently, as one technique to perform a suitable projection of the images, we have reported a projecting process of holographic 3-D images onto the underwater micro-bubble screen. It is easy to obtain Micro-bubble continuously with the help of transparent water tank, and therefore, it enables us to use it as a longtime stable screen. However, it rarely happens that available properties of the screen are not to be fully prepared on account of insufficient adjustment of the bubbling mechanism. As this reason, an improved bubbling system and its appropriate adjustment process have been strongly required. In this paper, we shall propose an improved, somewhat a new micro-bubble system and its adjustment process, and give some results studied about them. As this result, we found that this process is adjustable more easier than before, and gives us more stable, improved projected images than the ones obtained by previous system.

View full abstract

We propose a new denoising algorithm for medical ultrasound imaging. It is desirable to minimize speckle noise while maintaining a natural appearance for reliability improvement of diagnostic imaging. However, the high denoising performance wasn’t achieved because the decomposition of the noise and signal component was insufficient by a conventional harmonic analysis. To address this problem, we propose the denoising algorithm by using a multi-link wavelet, which is an analysis method for a directed acyclic graph (DAG) structure data, with an adaptive pixel sampling according to signal patterns. We performed comparative experiments of the proposed method and Grouplet known as conventional harmonic analysis. A high-performance was achieved in the proposed method in terms of a coexistence of noise reduction and signal preservation.

View full abstract

TV manufacturers produce TV sets equipped with image-processing technologies, such as super-resolution (SR), and frame interpolation to improve image and video quality. However, there is no evidence of the capability of the technologies on the TV sets, and the assessment of them is required. The assessment requires using multiple displays since we cannot take any signals after image processing from the TV sets. However, no method is adaptable to such the assessment. In this paper, a subjective assessment method, which is a combination of BT.500 and Scheffe’s paired comparison, is proposed and the subjective assessment experiments of 4K TV sets equipped with SR are conducted to confirm the reproducibility of the proposed method. The results prove the significant differences in the resolution of the displays, indicating that the proposed method has reproducibility and is useful for assessing the capability of SR implemented into displays.

View full abstract

As an effective reconstructing technique for the object characterized with multiple points, multiplex time-shared recovering process utilizing the segmented hologram elements is known. In order to make a colored reconstruction using this process, multiplex time-shared holographic reconstruction with the use of three color laser lights of red, green and blue may be strongly required. In this report, as a first step of this process, we have studied image characteristics attained by time-shared holographic reconstruction using blue laser light. As this result, although the shorter the wave length of reproducing laser light becomes, the severer the critical conditions of the required number of segmented points must be, we have seen that within the recoverable range of holographic images, relatively nice recovered images are come to be observed by adjusting the separation and the number of segmented points properly.

View full abstract

Users of smartphones and/or tablet computers browse and download confidential document files routinely. Therefore, the higher security level is needed for smartphones and tablet computers than conventional mobile phones (feature phones). From this kind of background, an image-based user authentication method for touch screen devices by using images shot by user oneself and gotten from the WEB, called SWIPASS, has been proposed. SWIPASS has resistance to observation attack, recording attack, and smudge attack, the most serious threats for touch screen devices. However, there is an issue such that when there exist images similar to the pass image in the terminal, legitimate users do not authenticate properly because they cannot recognize the pass image accurately. Then, in this study, we propose a method to eliminate images whose likelihood of being recognized as the pass image wrongly are high from candidates to be displayed as a decoy image. We moreover conduct an experiment to verify the effectiveness of the proposed method.

View full abstract

This research is to reveal what kind of information is effective for more precise reproduction of digitally archived historic buildings. This research examines two CG Computer Graphics reproduction projects of Mitsubishi Ichigokan (Marunouchi, Chiyoda-Ku, Tokyo), which has already been demolished and digitally reproduced; and The Clinical Research Building, the University of Tokyo Hospital (Hongo, Bunkyo-Ku, Tokyo), which still remains with its demolition scheduled soon. While for the former subject, we only had its drawing for design, photos, reference materials, and hearings with the person in charge of an architectural design office, which reconstructed the building, we were able to study the actual building for the latter, collecting the necessary information in addition to its drawing for design and photos. Our research leads to finding that the vital element for more precise CG reproduction is Detail Information and Material Information. This paper illustrates the information that would be effective in advanced CG reproduction, describing and contrasting the working details for both buildings.

View full abstract

Eaves Projection Mapping is the projection mapping to look up the eaves utilizing the structure of the house. In Takaoka Yamachosuji Dozotsukuri Festival 2015, this projection mapping was perfomed without destroying the atmosphere in streets. The projection area is six meters by one meter with the original frescos. This long shape projection used the projectors of one single focus and two small sizes. The contents are four construction done in the motif of Takaoka in Toyama. Music chose something conscious of Mikurumayama which is registered one as UNESCO intangible cultural heritage. Enjoyment method prepared this time is the shape that an audience will look up and lie on the back while standing up. It was also pointed out that it’s difficult for the audience to find the projection mapping being took place from the streat because that we tried not to disturb the atmosphere.

View full abstract

My laboratory is engaged in the research and development of the application development technology based on the concept that is “not depending on the others” by fusing with IoT technology and artificial intelligence. In this paper, I will introduce the latest results of our research such as the health care system using T shirts-type wearable device for a working woman without depending on a doctor, the social media intermediation robot that enables elderly to do daily work based on the interactive communication through SNS without depending on a care worker, and the barrier-free street viewing system that can be used for a person with a disability of wheelchair use without depending on an administration. Finally, we discuss the essence of IoT technology realizing “not depending on the other”.

View full abstract