We have developed an invisible code for smart phones that consists of high-frequency blocks and low frequency blocks. Finder patterns composed of high-frequency and low-frequency blocks are placed around the corners of the code to enable the code to be embedded in several places. Two key technologies were devised to implement this idea. One is high-frequency blocks composed of bright and dark elements in a checkerboard pattern. The other is a method of reading the code that employs edge detection to extract the code from a photograph of an image in which the code is embedded. To verify the effectiveness of our invisible code, we estimated the reading accuracy for code embedded in four common types of images. The results demonstrate 100% of accuracy could be achieved at distance of 100-150 mm.

View full abstract

In recent years, contrast improvement methods have been important image processing according to the spread of the products with camera. When the contrast of an image of photography is not good, a method to improve contrast by software is effective. The histogram equalization has been proposed as the typical conventional contrast improvement methods. This method generates an output image with a higher contrast because the output image has the maximum entropy. However, noises and intensive improvement may occur, and a resultant image may become unnatural. The purpose of this research is to propose a contrast improvement method that uses the feature analysis of an input image. Two image features are used in this method. First, the detail variance and the background variance are used to reduce noises and to set a clipping level. Second, the sharpness factor by the delta histogram is used to adjust improvement effect and to generate good contrast images. The experimental results show that the images that had good and natural contrast were generated by the proposed method.

View full abstract

This paper proposes a feature selection method that improves the recognition rate significantly for not only training samples but also unknown samples by using the principle of margin-maximization in the support vector machine (SVM). SVM is well-known as a recognition method that can discriminate unknown samples with high precision, so feature selection with high recognition rates for unknown samples can be expected by adopting the principle of margin-maximization, the technical basis of SVM. We perform experiments on five sets of hand-written Kanji character patterns. Each set consists of patterns in two similar categories. The results show that the proposed method improves the recognition rate significantly for not only training samples but also the unknown samples as expected.

View full abstract

Object tracking is a key objective of research in image engineering and considerable research has been directed toward developing tracking techniques based on particle filtering, which generally utilizes similarities in color, shape, texture, and other attributes of a target in likelihood calculations. However, partial occlusion of the target tends to reduce the accuracy of target recognition and is thus a significant obstacle to effective tracking. We therefore propose a technique for robust object tracking even when the target is partially occluded. It employs hierarchical likelihood calculations in particle filtering.

View full abstract

In this paper, we propose a method to track multiple players and ball in a football video which is captured by monocular camera. In the case of capturing the video from a single view, there might be a lot of occluded situations of players. To overcome this problem, we propose a robust tracking method for multiple players by combining Particle Filter and Classifier. In this method, we apply Homography to extract player's position on the pitch.

View full abstract

Blood vessels visualization in brain region is an important prerequisite for a number of clinical procedures such as diagnosis, image-guided neurosurgery, and pre-surgical planning. At the clinical scene, it has been used to MIP (maximum intensity projection) method for the MRA image. The MIP is useful method, however MIP has a problem that the vessels are hidden by non-brain tissues like skin and fat structures.Then, these unnecessary parts have been manually deleted by operators, and the workload of this task has become a great burden for the operators. In order to the above problem, this paper presents a simple and fast brain segmentation method for the blood vessels visualization. Our method has attention to the structural feature of the brain MRA image, and the full-automatic skull stripping is realized using Minimum Intensity Projection(MinIP) and the region growing method. In our experiments, the proposed method is evaluated with clinical datasets. As the results, it is seen that our method provides sufficient accuracy for clinical use, and the computational costs of our method is about half compared with the manual work by operators.

View full abstract

In this paper, we propose a new scheme of time-diversity system using optical code division multiplexing (OCDM) in free space optical (FSO)communication. Our design is a full-optical communication system without optical and electrical conversion that facilitates synthesizing the time-diversity signals. We found that employment of the proposed scheme can effectively minimize the influence of scintillation caused by atmospheric turbulence. We analyze the bit error rate (BER) performance of the time-diversity scheme through atmospheric turbulence channels and its efficiency enhancement with a digital TV signal. It not only considers the intensity scintillation of atmospheric turbulence, but also considers scintillation temporal -frequency. We present the theoretical analysis which is based on lognormal models, and validate the performance of the proposed scheme through simulations. Simulation results are further demonstrated to confirm the analytical results. They show that under some circumstances the bit error rate can be improved by a few orders of magnitude using the time-diversity scheme to compensate for the scintillation.

View full abstract

In this paper, we describe the method to archive the mural paintingsin the Royal tomb of Amenophis III as digital data. We took 1000 to 2000 photographs per wall with the single-lens reflex digital camera and the equipment of automatically taking panorama photography. To stitch these photographs we used software on the market and the originally developed one.The digital images of the mural paintings obtained by this method have the sufficient quality for researchers to observe, whereas the cost for digitizing is reasonable because we used only conventional equipment.

View full abstract