Search

The Drosophila visual center shows columnar structures, basic structural and functional units of the brain, that are shared with the mammalian cerebral cortex. Visual information received in the ommatidia in the compound eye is transmitted to the columns in the brain. However, the developmental mechanisms of column formation are largely unknown. The Irre Cell Recognition Module (IRM) proteins are a family of immunoglobulin cell adhesion molecules. The four Drosophila IRM proteins are localized to the developing columns, the structure of which is affected in IRM mutants, suggesting that IRM proteins are essential for column formation. Since IRM proteins are cell adhesion molecules, they may regulate cell adhesion between columnar neurons. To test this possibility, we specifically knocked down IRM genes in columnar neurons and examined the defects in column formation. We developed a system that automatically extracts the individual column images and quantifies the column shape. Using this system, we demonstrated that IRM genes play critical roles in regulating column shape in a core columnar neuron, Mi1. We also show that their expression in the other columnar neurons, Mi4 and T4/5, is essential, suggesting that the interactions between IRM proteins and multiple neurons shape the columns in the fly brain.

Digital watermarking is the main method by which prevent illegal copying of proprietary images. In addition, to reduce the amount of damage caused by illegal distribution of proprietary images, a fast search technique for illegally distributed images on the Internet is necessary. We have proposed a two-stage search method for digital watermarked images. In the present paper, we propose a new high-efficiency pre-search algorithm for proprietary images that are suspected of being copied and distributed on the Internet. The proposed algorithm is capable of quickly identifying illegal copies from large image groups using mainly the correlation coefficient of compressed representations between the original image and a target image by convolutional autoencoder. The proposed search method has tolerance for attacks such as scaling, image compression, color reduction, and intensity conversion. The performance of the proposed search algorithm is evaluated through computer simulations.

著者人数: 30名, Accepted: 2018-11-09, 資料番号: SA1180263000

For real-time applications such as autonomous driving and virtual reality (VR), we previously proposed an ultra-low-latency video coding method, which adopts line-based processing for Full-HD video. In this paper, we newly propose a design and implementation of the ultra-low-latency video encoder. In order to reduce the hardware amount, image-prediction specification is optimized for our previous work. Applying a high-level synthesis (HLS) design methodology for Xilinx FPGA, the implementation results of logic count with 10,677 LUTs, 3,714FFs and 66 DSPs is obtained. The implemented video encoder achieves less than 1.0 µs low-latency and compression to 39.4% without significant visual degradation. As a result, cost effective ultra-low-latency video encoder is implemented for low cost FPGA.

This paper describes a scalable architecture for high-resolution real-time optical flow processor and the design for 60-fps real-time Full-HD system. A previously proposed 30-fps WXGA real-time processor is re-designed as an IP (WXGA-IP) for the proposed Full-HD processor, which adopts a scalable multiprocessor structure. In addition, three methods are newly proposed, namely an improved hierarchical method, an improved initial value creation method, which reduces the number of iterative computations, and an adaptive acceleration parameter selection method that uses the built-in CPU. The realtime processing of 60-fps Full-HD video, which requires four times the processing power of WXGA video, is realized using twice the hardware while maintaining flow detection accuracy at a similar operating frequency compared to those of WXGA-IP.

A one-to-one correspondence between in-situ chorus elements near the equator and rapid luminosity variations of a pulsating auroral patch on the conjugate ground site (Gakona, Alaska) is identified from coordinated observations by the Arase satellite [1] and a ground observation network of PWING (study of dynamical variation of Particles and Waves in the INner magnetosphere using Ground-based network observations) [2]. The spatiotemporal development of the pulsating auroral patch was captured with a 100-Hz sampling imager. The temporal variations of a pulsating auroral patch are associated with not only a train of chorus elements near the equator, but also possibly subpacket structures of the chorus elements with tens of milliseconds. Here we found that the luminosity variations in the well-known few Hz range were observed only at the edge part of the auroral patch in this event, because they were caused by the spatial expansion and contraction of the auroral patch from the center location. The spatial size at the ionosphere (altitude of 110 km) of the auroral patch was approximately 15-km long at the short axis in the magnetic north-south direction and 70-km long at the long axis in the east-west direction of an ellipse. The spatial scale projecting the ionospheric auroral patch to the equatorial plane of the magnetosphere was roughly a circle of 900 km in diameter, which is comparable with a small spatial scale of relativistic electron microbursts. These observations suggest that the pulsating auroral patch associated with chorus elements reflects not only a temporal variation of wave–particle interaction in the magnetosphere, but also a transverse spatial profile (across the geomagnetic field line) of wave–particle interaction. In this presentation, we will present the spatiotemporal variation of the pulsating auroral patch associated with chorus elements in detail.

Columns are structural and functional units of the brain. However, the mechanism of column formation remains unclear. The medulla of the fly visual center shares features with the mammalian cerebral cortex, such as columnar and layered structures, and provides a good opportunity to study the mechanisms of column formation. Column formation is initiated by three core neurons in the medulla, namely, Mi1, R8, and R7. The proper orientation of neurons is required for the orientation and arrangement of multiple columns. Their orientations may be under the control of planar cell polarity (PCP) signaling, because it is known to regulate the orientation of cells in two-dimensional tissue structures. In this study, we demonstrate that the ligands DWnt4 and DWnt10 expressed specifically in the ventral medulla and dorsal medulla, respectively, globally regulate the columnar arrangement and orientation of Mi1 and R8 terminals through Fz2/PCP signaling in a three-dimensional space.

There is a similar image retrieval method which uses a normalized histogram of the pixel values of the original image and that of the pixel values of the illegally copied search target image. In this method, when the correlation coefficient between the two histograms is greater than a predetermined threshold value, the target image is to determine to be similar to the original image. However, the contents of the image that has received ornament attack is different from the contents of the original image, this means that it is searching for images similar to the original image is very difficult. In general, ornament processing of the image is applied to the peripheral region of the image. Therefore, if only the pixels to limit in the ROI (region of interest) which are not affected by the ornament attack are used to create the normalized histogram, it may be effective. We, in order to extract only the pixels included in the ROI, have proposed a method of using a mask image. In this paper, it describes some detailed study about the element techniques including using the mask image and normalization histogram. Moreover, the computer simulations show the effectiveness of these element techniques.

Recent years have witnessed a drastic rise in the number of image content on the Internet. In this context, several digital watermarking techniques have been investigated in order to prevent illegal duplication and distribution of such content. Illegal users in turn attempt to remove digital watermarks from such content, and collusion-attacking methods, such as an average value attack, constitute a serious threat to digital watermarks. In previous research, we proposed a restraint scheme of average value attacks, which is one of collusion attacking. By the scheme, a region of interest (ROI) on the image is translated based on seam carving technique. When the image content is done by average value attack, a reconstructed image quality is decreased. However, the ROI is located manually on the content. Therefore, we applied a saliency map to the scheme for locating ROI automatically. Experiments have been shown good results.

A 100-MHz 51.2-Gb/s packet lookup engine LSI for packet inspection was proposed in a previous paper. High throughput and low power consumption was realized by introducing a combined mismatch detection circuit and linked-list hash table, minimizing the amount of memory accesses. The LSI has an automatic table update function. In this paper, this automatic rule registration and deletion function on the packet lookup engine LSI is described, meaning various tables included in the circuits, that is, the Mismatch Table, Index Table, and Rule Table are automatically configured by only inputting rules to be registered or deleted using embedded hardware on the LSI. This function utilizes a match and mismatch assessment for normal packet inspection operations.