This paper presents a new watermarking method based on fractal image coding. Fractal image coding is first carried out where fractal codes are divided into two ranges related to watermark information 1 and 0. Then fractal decoding is performed to obtain an image embedded with the watermark information. For the extraction of watermark information, the image is fractal-coded to decide the range of fractal codes. The fractal codes employed for watermarking are luminance scaling, isometric conversions, and horizontal/vertical shift of a domain-block. It has been examined how image qualities are affected by the watermarking process and to what extent watermark extraction accuracies are degraded by quantization attacks.
In this paper, we propose a hybrid type digital watermarking method which applied two kinds of technique. Experimental results show that the watermark embedded by the proposed method is perceptually invisible and robust against various attacks such as JPEG compression and geometric transformation.
Lossy coding of error diffused images is an interesting topic but has not been studied well yet. In this paper we study two different categories. First one is to use multi-level lossy image coding methods such as JPEG, JPEG-LS and JPEG2000 including the possibility of reduction and enlargement, with the error diffusion processing at the receiver. The other category is to introduce the loss by sending the number of black pixels in plural pixels, where we studied the simplest case of three levels in two pixels. In the former category, we found that JPEG2000 achieves best, and the error-diffused image at the receiver is compared with the restored bi-level image of the latter by the SNR of inverse halftoned images. The result shows that the first category works better for the low-frequency dominated images and the second one works better for the high-frequency dominated images.
A fidelity color calibration method was developed using the colorimetric characterization of an inkjet printer in order to realize colorimetric reproduction. To make a prediction the value of RGB to calibrate the input of printer drive, the measurement XYZ data of a printer was used to estimate RGB from XYZ by prism interpolation method. The XYZ of input color were compared with those of the values capture by the printer. The average color difference, ΔE, between the input and printed one were 1.89.
This paper proposes a new approach for precisely and stably detecting changes in image scaling factor. DFT amplitudes for reference and input images are first transformed to the polar expressions. In the transformation, the cause of detection errors is analyzed and optimal frequency filtering for reducing errors is found. The position shift between the polar spectrum is detected by our PSA method. The weight function effective for accurate detection of the shift is discussed. The approach is applied for detecting changes of image scaling for four typical types of images. It becomes clear that the scaling factor can be detected with higher accuracy than that of the conventional method. The accuracy dependences on the patterns and scaling factor are also made clear and discussed.
Color histogram which shows the frequency of RGB values of pixels of a color image is used for image processing widely. We give the frequency information and the position information of pixel to each point plotted to RGB color histogram space, and defined as the frequency image what expressed the frequency information visually on the image, and have proposed as a new feature extraction method of a color image. This frequency image is useful to development of many image-processing techniques. In this paper, we outlines about the creating method of frequency image and its applications.
This research aims to extract persons from TV program and to measure the appearance rate. In order to achieve person's identification without depending on the direction of the face, the face is detected using Haar-Like feature, then, the person's clothes color is acquired based on the position of the detected face. The color histogram of the clothes is registered as a dictionary, and then clothes color of the succeeding scene is compared with this dictionary. After all processing is finished, and rate of the appearance of the specified person is calculated. Finally, rate of appearance in processing is compared with actual rate of appearance, and the extraction rate shows very high.
This paper shows that a method of generating a three-dimensional scenery map by removing shadows in an aerial photograph, and combining corresponding three-dimensional map, in order to realize a three-dimensional map that does not depend on the season and the time. First, the features of the sunny area and the shadow area are investigated. After that, the aerial photograph is classified into road area, roof area, and other, and the shadow extraction and the shadow correction processing are achieved for each area. Finally, an aerial photograph without shadows can be obtained by composing three areas, so that a three-dimensional map with shadows at arbitrary season and time can be generated.
Southern Ishikawa area research unit, which has been selected as one of the promotion of science and technology on city area project, is developing a visual simulation method for traditional craft materials. This paper reports a method for representing gold leaves faithfully using Bidirectional Reflectance Distribution Function (BRDF) and surface bump data of materials. In this method, the color variation to gold leaf by the angle of incidence of light is confirmed and it can be perfectly reproduced. In addition, a peculiar wrinkle to gold leaf is also expressed.
This technical report provides experimental results of center-view image generation from a circular array of real captured images. In this experiment, our previously proposed method was applied using estimated camera calibration parameters. The experimental results show that artifacts due to camera calibration error and noise were not visibly observed; hence our method effectively works for real images.
In this paper, we improved Image Retrieval System of Flowers that can be easily used in a mobile environment.On the limited specification of the mobile phone, we improved the technique to extract a flower region from a source flower image and improved the user interface. In extracting a flower region, we solved the failure of painting out processing, miss-extraction of rear flowers, and incompleteness of color specification method. In the user interface, we proposed a new color specification method and a way to confirm the result of the extracted region of a flower. By experimenting on the flower region extraction using our method, the result shows that the rate of the succeeded extractions was 97.1% and increased 21.3 points compared with the previous system.