Line drawing style Ascii-art is one of the method to express images with shape of characters. The Ascii-arts are conventionaly made by manualy, but then those made by automaticaly, too. In this study, we thought that one to many relationship of exchange line-art and character can make redundancy to Ascii-arts. This report presents an auto-making method of line drawing style Ascii-arts that based on our idea. From experimental results, we revealed that matching parameter of pattern-matching can control quality of Ascii-arts and information hiding capacity. And we found availability and disability of bilateral filter for clean abstraction of outline with natural image.

View full abstract

In this paper, we propose a new information embedding method which checks whether the error-diffused pixel value is within the specified range. Our experimental results using the images Lena and Baboon, and using the high-resolution standard color test image N1 (for the C, M and Y components excluding the Black component) show that information can be embedded in about half of the total pixels by keeping the MAE value of an image with no information embedded, when the error-diffusion filter coefficients proposed by Floyd and Steinberg are used, and in 60 to 70 percentage with less degradation of image quality when the filter coefficients proposed by Jarvis, et al. are used. This means that much additional information can be embedded in the binary error-diffused dither images for printing. The information embedded by this proposed method is classified as watermarks, not steganography, because the information embedded is extracted with the original multitone image.

View full abstract

In this paper, we propose a real-time object detection LSI for mobile videophone. The LSI achieves real-time processing speed by stereo vision algorithm using temporal data relationship between frames. The algorithm uses a last frame result to reduce the detection area and search range to the edge of the object and the background. We have designed object detection LSI by using this algorithm. The LSI can process QCIF stereo images 30 frames/sec and logic area is 15182 gates.

View full abstract

Pixel mixture is a technique to reduce read-out time of an image by mixing multiple pixel values on an imager. Values of same color pixels are mixed by ordinary pixel mixture methods. However, the same color pixels are not neighboring on widely-used Bayer pattern. This degrades the resolution of the resultant image. In this paper, we propose a neighbor pixel mixture method. The proposed method mixes neighboring pixels by using a linear combination. We also discuss a guideline for designing the coefficients of the linear combination.

View full abstract

This paper presents a new quantization technique based on the characteristics of graycode, and a new technique of lossy image coding, which is composed of predictive coding, wavelet transform and the quantization technique. Graycode is a code in which codes for two natural numbers whose difference equals one differ only at one bit from each other. When a bit of graycode of a value is changed, the average of the original value and the changed value can be represented by more significant bits above the changed bit. In the quantization technique, values to be encoded are quantized so that their acceptable quantization error ranges are fixed. The quantized values are encoded with quadtree and arithmetic coding. Experiments were performed in which the proposed technique was compared with JPEG and JPEG2000 in respect of compression ratio and image quality. It was shown that the proposed technique gives better quality than JPEG when the predictive coding is employed. If the target quality of images is high, the proposed technique gives equal or better quality compared to JPEG2000 when the wavelet transform is also used.

View full abstract

Vector quantization (VQ) is one of image compression techniques. VQ has been improved by using a generalized harmonic analysis (GHA) that has a good performance in analyzing and extracting a low frequency component. The technique can be used only on VQ with a small codebook because it generates one codebook for each image. This paper proposes to use a fixed codebook for the technique. In this case, the proposed technique uses a GHA for different image at codebook making and image coding. By using a GHA and VQ together, the proposed technique obtains a better performance than VQ. And then, the proposed technique is evaluated. The result of evaluation shows that the proposed technique has 0.1[dB] to 0.3[dB] better image quality than VQ.

View full abstract

For the future television system, multi-view video coding is under the standardization in International Standardization Organization, which is used for arbitrary view-point image synthesis or volumetric image presentation. Conventional motion compensated frame prediction can be applied to the spatial prediction between camera images. If the camera arrangement is known, a faster and more accurate prediction will be feasible. In this paper, assuming the object face direction, the relationship between camera images is analyzed and their utilization for block displacement and deformation prediction is investigated. Preliminary experiments showed bi-directional prediction, variable block size prediction and reference block shape compensation are suitable for disparity compensation. The possibility of block displacement vector compression and block shape information compression is also investigated utilizing the relationship between them.

View full abstract

Since motion in an image sequence is non-stationary, perceived image quality can be improved by an adaptive frame rate control in accordance with statistics on its motion when the sequence is compressed and transmitted in a band-width-limited network. The authors have proposed an adaptive frame insertion technique for image sequences. However, the technique often gives motion jerkiness at a frame-rate transition point in an almost constant-motion section, which requires further improvement. This paper thus proposes a new adaptive frame interval control technique that gives high subjective image quality. A MOS estimation technique is also presented by investigating the relation between the subjective image quality of resulting sequences and their average frame rate. The effectiveness of the proposed technique is finally evaluated through some experimental results.

View full abstract

Recently, the infrastructure maintenance to achieve an ubiquitous environment progresses steadily. In this environment, it is expected that the use of a wireless communication increases because the equipments installation is easy. The error occurs frequently in a wireless communication compared with a conventional wire communication because of interference and the influence of the obstacle. Moreover, the frequent route cutting caused by moving of the mobile terminal in the ubiquitous environment. In this study, it is aimed at the construction of the method to obtain image transmission results which has visual stable quality. In this paper, we focused attention on the image transmission method by using multiple routes, and proposed a split method that splits luminance and chrominance independently by considering visual characteristics. The proposed method is simple operation and can flexibly change the redundancy. The effectiveness of the proposed method was confirmed by the simulation. Also, we evaluated simultaneous communication by performing communication experiments at the packet level.

View full abstract

In this paper, we introduce a novel post-processing method that reduces image degradation caused by lossy image coding methods. Main problem of conventional post-processing methods is its heavy computational load because they are using iterative calculations including DCT and IDCT operations. On the other hand, we propose a filtering method that estimates quantization error in the transformed domain that does not require iterative procedures. Also we introduce a simple adaptive filtering method for better estimation of quantized noise. By the decoding examination, we found that the proposed method provides quite stable image quality in all bitrate regions in comparison to a conventional method.

View full abstract

Today, the opportunity to appreciate visual contents on the device of various sizes, such as an large theater, a cellular phone, etc. has increased. In addition, contents are appreciated in various different environments, such as in the train, and the outdoors, etc. However, the impression from the same image may differ in the different environment. In order to solve the problem, the technique of the color production in consideration of environment and KANSEI is required.
In order to evaluate the relation between the sizes and the colors, we investigate KANSEI evaluation of images by changing the reproduction color and display size. In addition, we investigate the relation between stimulus words to identify the evaluation structure on image contents.

View full abstract

Various methods have been proposed for Inverse Halftoning (IH) of Error Diffusion (ED) Images. Among them Table Look-Up method based on every possible combination of the bi-level output of reference pixels has been considered to be most effective and quite practical since IH function can be generated by simple averaging of the histogram of original con-tone levels of the target pixels in each context, and the image quality can be controlled by the number of reference pixels. In this paper, we will optimize the template and propose a hybrid system, in which each context is composed of the combination of the conventional logic of some part of reference pixels and reduced state information given by the sum of black pixels in other reference pixels. The result showed that the SNR compared under the same number of contexts are found to be improved from the conventional method especially when applied to the images not included in the training data.

View full abstract

Efficient Inverse Halftoning (IH) methods of Ordered Dithering Images were studied for clustered dot type and dispersed dot types. They are based on the Table Look-Up method that utilizes every possible combination of the bi-level output of reference pixels, which has been considered to be most effective and quite practical. In this paper, we will optimize the reference pixel template with a new introduction of flat-area detection and dither matrix address information. The flat area detection mechanism will contribute to save the memory capacity by enabling to classify the reference pixel template according to flat area or non-flat area. The dither matrix address information is found to contribute to improve SNR in some memory size condition. With the proposed ideas, the memory size to provide the same SNR in the conventional method is reduced to between 1/8 and 1/128, and generality of designed function was also found to be improved.

View full abstract

Recently, biometric person authentication has been in the spotlight. However, in conventional biometric person authentication systems, each individual's template is stored as it is in the system. Hence, when a registered template is not properly protected, the risk of impersonation using biometric information restored due to template leakage by a third party arises. In this paper we propose a method whereby template information is either split or partly deleted, and restoration of the template by the legitimate individual is made possible by not using the registered template information only, but rather using error-correcting code. The proposed method can be applied to a general biometric authentication system. In this paper the sacurity of templates is evaluated by the simulation of a speaker verification system.

View full abstract

A design method of non-periodic halftone screen for color printing to reduce moire and grainness and to improve the imaging stability is proposed. With the irregular dot formation and the difference of the spatial frequency character between the halftone screens for each color separations,moire and grainness caused by superimposed each halftone screens are suppressed. Additionally, the halftone dot for the black screen is coarser than that for other color screen. Because the visual density of the black solid area is the highest among the four process color inks, visible banding pattern caused by scan imaging is also efficiently suppressed.

View full abstract

A method to acquire an pan focused image is invesgated by using a few differently-focused images. According to this method, the pan focused image is composed by selecting a focused part from two images and combining it with the rest. Here the unevenness coefficient regularized for image brightness is used to determine the focused part. In this paper, we show that by this method we can discriminate between the focused part and the unclear focused part in the image using the unevenness coefficient. Our computer simulation can successfully demonstrate that this new technique provide effective pan focused image.

View full abstract

We study about an object tracking method based on Particle Filter for a moving camera. To continue target tracking, updating of the reference histogram and clarify the separation of target / background is important. In this paper, we use a spatial distribution of particles (likelihood distribution) for updating the reference. And we add a histogram feature near the target (around reference) when likelihood calculation for clarify the separation. Actually, as a result of the experiment about any kinds of videos, it succeeds in tracking of each target, which includes a non-rigid target (flag).

View full abstract