As the need to use various devices increases, we need to produce the compressed image at different resolutions. The development of an image compression format that compression and resolution conversion can be performed simultaneously is desired. In this research, we suggest a lossless image compression format called “Function Image Compression (FIC)". Since FIC has function representation of an image data, resolution conversion can be easily performed because resampling is easy once it performs function representation. We show the results of zooming images, calculation costs for resolution conversion, and image compression ratios for the evaluation of FIC system.

View full abstract

Due to the advances in computer performance and networks, highly detailed models in industry and entertainment are widely used. Compression is essential for the environmental load on transmission, storage, and display resources. The main contribution of this paper is to compress implicit function representation by using CSRBF after mesh simplification is performed. Our coder can improve the compression performance without using mesh connectivity unlike other conventional techniques. Our technique has another advantage to be able to use characteristics of implicit function directly. In the compression ratio, our results demonstrate that the proposed approach has better image quality and error measurement than conventional techniques.

View full abstract

A lot of collision detection algorithms for three-dimensional space objects have been studied. At present, they are important methods in wide area, such as CAD, computational geometry, and computer graphics (CG). It is still difficult, however, to detect exact collision faces quickly because of high calculation costs. In this paper, we propose an algorithm of exact collision detection. In our algorithm, we introduce generation of Layered Boundary Spheres (LBS) to select candidates of collision faces quickly, so that exact collision faces are detected. Additionally, our method is applied to PC cluster environment. PC cluster enables discrete collision detection, which decreases calculation cost of each computer. Therefore it will be possible to obtain exact collision faces quickly.

View full abstract

Many optical reflection models have been proposed in computer-vison and computer-graphics. Parameters contained in the models are frequently determined in empirical ways. High sensitivity of data to experimental conditions makes it difficult to formulate the parameters. In order to compare the results between many experiments, a generally usable standard of radiance must be extracted in each of experiment. We proposed a method of estimating normal diffuse radiance (NDR) of cuboids' surface, i.e., radiance of a normally illuminated surface, without accurately adjusting the surface direction toward ligth source. This paper deals with a reflection model at a rounded edge of cuboids by using the NDR as a standard of radiance and gets following results. (1) The diffuse radiance at a rounded edge is well approximated by a function of cos² φ where φ is an incident angle. (2) By using the NDR, a relation between diffuse radiance of an edge and those of surfaces of cuboids can be quantitatively formulated. (3) Theoretical results (1) and (2) are examined by experiments and are validated.

View full abstract

In developing the flight data recorder for small helicopters, no change in the body is desirable because of the flight regulations. Thus we propose the method to obtain the readings from video-taped images of the instrument panel using image processing techniques. This paper describes the methods to read the helicopter instruments by a computer. The instruments with an indicator and a scale could be read by the typical procedure for the image processing. However it did not work for the reading of the attitude indicator. For attitude indicator, we proposed the method based on the model based approach which calculated the projections of its shape for given pitch and roll angles: then searched for the combination of these angles that give the projection matched best to the input image. The experimental results showed that the accuracy of the readings obtained by the method almost met the requirement.

View full abstract

Relevance feedback has been shown to be a very effective tool for enhancing retrieval results in text retrieval. In recent years, the relevance feedback scheme has been applied to Content-Based Image Retrieval (CBIR) and effective results have been obtained. In the relevance feedback, it is important that how to reflect a user's intention to the system. As for those methods, a burden to a user was light, but there was a problem that it was hard to feedback how the result was irrelevance. This paper realizes a relevance feedback to reflect visually how the result was irrelevance without giving an excessive burden for a user. We propose a CBIR algorithm using Self-Organizing Map (SOM) with visual relevance feedback scheme. The relevance feedback is performed by moving an irrelevant image to a relevant position by a user. We show the effectiveness of the proposed technique by subjective evaluation experiments.

View full abstract