Center/Surround (C/S) model of human vision has hinted to improve the sharpness, lightness, and contrast. Retinex is a root of C/S model which restores the reflectance image by removing the spatial non-uniformities of illumination. To recreate a viewer's sensation of realistic scenes, visual Tone Mapping Operators (TMO) have been actively developed during the past decade in two ways of Spatially-invariant and Spatially-variant approaches and now the latter is major stream. However, since most of spatially-variant TMOs are designed independent of the Tone Reproduction Curve (TRC) provided by the system, they meet a difficulty in keeping the visual local contrast. This paper proposes an improved spatially-variant TMO by an “area-based” C/S model that maintains local visual contrast based on Weber's law under the given TRC. The proposed TMO purely improves the visual contrast better than our previous “pixel-based” model without changing the lightness and colorfulness. But for the poor quality images taken under bad illumination or low dynamic range camera, the model recreates the better image appearances in color-tone as well as contrast in cooperation with an adaptive Multi-Scale Retinex. The paper introduces the experimental results applied to natural images in comparison with our previous and other existing models.

View full abstract

The error diffusion method creates high-quality dithered pictures. It can be expanded to multilevel dither processing. However, the multilevel error diffusion method causes picture quality deterioration around a middle quantization level. Some studies are reported to solve this problem. However, the previous methods have complicated algorithm and may not be the most suitable method. This paper shows that both black pixels and white pixels should be existed around the middle quantization level, and proposes a new three level error diffusion method with a very simple algorithm. It can be achieved by restricting the quantization condition. Additionally, picture quality can be improved by limiting the restriction target to where the input levels are around the middle quantization level.

View full abstract

Our aim is to produce a holographic 3D TV system adopted a portable terminal information processor of the small size, just like a cellar phone. In fact, as for a cellar phone, since only a small number of pixels are permitted for displaying, it seems to be valuable to study a holographic 3D reconstructing process adopted such devices in which small numbers of pixels are available. In this report, we studied reproducing process employed a technique which enables us to record the object with many points and to reconstruct it using time sharing methods, and we present here a technical process of the improving characteristics of the reconstructed images produced from the objects with complicated forms.

View full abstract

It is known that if a lens-array sheet is put over periodic pattern and those periods are close to each other, they make magnified images with depth by moire interference. Using this principle, we contrived some methods to display periodic stereo images with 3D parallax by making periodic images from images captured from rotated object by fixed angles. By the experiments using lenticular lens and square lens array sheet, it was confirmed that expected function was realized. In addition, we studied some periodic stereo images made from some lissajoues figures.

View full abstract

Recently, several digital watermarking techniques have been proposed for hiding data, such as the copy right information, in a frequency domain of image signals to protect the copy right of the content. However, in order to detect a watermarking sufficiently after heavy compression, the quality on an image might be visibly deteriorated by a digital watermarking with strong intensity. We have proposed a verification method for a digital image file using a discrete wavelet transform. The proposed method uses elements of multi-resolution representation of an image for identification and never changes the image. The experimental results show that the proposed method has an excellent property on the tolerance to compression.

View full abstract

In this paper, we propose a novel multi stream decoder system that applies load adaptive decoder control method to facilitate efficient plural decoding processes. Conventional decoder control methods incur problems about processing efficiency and practicability to be implemented on an embedded system which deals with video contents provided by common video services such as digital TV broadcasting in the field of information appliances. We investigate the load adaptive decoder control method that adjusts the load of decoding process based on characteristics of video stream. The experimental result using HD MPEG-2 video streams shows that our method can suppress the frame drop rate and degradation of the image quality in multi stream decoding process as the number of streams increases and that there is little increase in processing time.

View full abstract

Recently, car vision technologies have been paid much attention in the field of ITS (Intelligent Transport System). In particular, techniques for automatic detection of pedestrians or humans have been studied enthusiastically. A robust human detection algorithm has been reported recently. The HOG feature proposed by Dalal and Triggs 1) is a well known feature for representing and recognizing a human image. But, in its original algorithm, it includes an unnecessary part of the image such as the background. In this paper, we propose a method of detecting a human and its body direction employing the HOG feature based on a human model. The reason for detecting body direction is to predict his/her sudden step into a road. To the best of our knowledge, the HOG feature has been employed to date only for detecting a human and not for recognizing its body direction. The proposed method is described in the following and the performance of the method is shown by the experiments employing real outdoor scenes.

View full abstract

View is interfered with a rain noise when we take a video at the time of the rain outdoors. In this short-paper, we show a method to interpolation a candidate rain pixel in an axial direction in time, and we get the removed image of the rain noise. I compare a minimum filter, a median filter, a spline interpolation as a method to interpolation for and show the effectiveness of the spline interpolation.

View full abstract