We have studied an optical interface for Super Hi-Vision production systems. In this study, we develop an optical interface which can connect full-resolution Super Hi-Vision production equipment by an optical fiber. It turns out that this interface conforms to the bit-error rate specifications of SMPTE 435 by a connection experiment.
We examined SRTS (Synchronous Residual Time Stamp) clock recovery method for received HD-SDI (High Definition Serial Digital Interface) signals over SONET (Synchronous Optical Networks) or SDH (Synchronous Digital Hierarchy) Networks. Experimental results showed that received HD-SDI signals satisfied jitter specification standardized by SMPTE.
For re-transmission of digital broadcasts over FTTH, we have proposed a baseband time-division multiplexing system and multi-channel re-modulation system. We produced an ISDB-T multi-channel re-modulator experimentally.
A new classification method for adaptive image processing is proposed. The method divides an image into some areas and classifies them into flat areas or complex areas based on those extent informations. Adaptive image processing using this method improves images better than past methods.
We are developing a new segmentation method for Texture images using Gaussian MRF model based on wavelet frame feature of images. In this paper, we improved the segmentation procedure based on the modeling. As a result, we obtained higher accuracy of the segmentation of the image including two texture classes.
We proposed an image segmentation method using texture features and Markov random field models in order to extract the contour of a particular segment. In this paper, we compared segmentation rates obtained by our methods the numbers of models are different among.
In this paper, we propose a novel self embedding watermarking scheme using block truncation coding (BTC). In our scheme, two components which are derived from BTC were utilized for embedding watermark. Computer simulations show that our scheme successfully detect modifications and recover them to some extent.
This paper presents a method to restore a high ISO-sensitivity color image. Our proposed approach is composed of three consecutive stages. At the first stage, observed raw color data mosaicked with the Bayer color filter array are demosaicked with the bilinear interpolation method. At the second stage, to remove observation noise selectively, we apply the l^1-l^2 color shrinkage to undecimated Haar wavelet coefficients of the primarily demosaicked image. At the third stage, we apply the interative color-image deblurring approach with the undecimated Haar wavelet transform, to the denoised color image.
The shrinkage operation in the wavelet transform domain is known to perform approximately like the TV (total variation) regularization in the image space domain. An image deblurring method with the classical shrinkage scheme in the wavelet transform domain was recently proposed by Chan et al. This paper proposes a new sharpening-demosaicing method using the redundant Haar wavelet transform and our previously proposed color shrinkage scheme. Our proposed method performs demosaicing and sharpening simultaneously. Simulation results demonstrate that our proposed method works as well as our previously proposed sharpening-demosaicing method using the TV regularization in the image space domain.
The color-imaging scheme with the 4-color CFA has the advantage of restoring high spatial-frequency components very successfully, but it has the disadvantage of poor color reproduction. To improve the color reproduction of the color-imaging scheme with the 4-color CFA, this paper proposes a new linear demosaicing method with the undecimated Haar wavelet transform, and demonstrates its superiority in color reproduction over the existing linear demosaicing methods.
Previously, we have proposed the l^1-l^2 color shrinkage scheme utilizing inter-channel cross-correlations. We have shown that the l^1-l^2 color shrinkage scheme is useful for denoising color images not only in the redundant wavelet transform domain but also in the redundant DCT domain. However, as for the color-image denoising in the redundant DCT domain, color smears often remain when the block size of the DCT is set small. On the other hand, noise in the vicinity of sharp color-edges is little removed when the block size of the DCT is set large. To alleviate these drawbacks, this paper introduces the redundant DCT of adaptive block-size into the color-image denoising method, and experimentally demonstrates its improvements.
Previously we have proposed a method to analyze a noisy image. The analysis method firstly applies the γ correction and the nonlinear intensity transformation to both a noisy image and its corresponding noise-free original image, decomposes each image into its three components with the BV-G image-decomposition method, and then computes the SSIM for each separated component between the noisy image and the original image. This paper extends the analysis method to an analysis method in the opponent color-space, known as the analysis color-space employed by the HVS, and applies the method to evaluations of denoised color images provided by the state-of-the-art color-image denoising methods.
Previously, we have proposed the three different l^p-l^2 color shrinkage schemes, for p=0, 1 and 2, utilizing inter-channel cross-correlations, and have shown that the l^1-l^2 color shrinkage and the l^1-l^2 color shrinkage improve performance of the color-image denoising in the redundant Haar wavelet domain. Superiority or inferiority in performance of the three color shrinkage schemes depends on prior distributions of true wavelet coefficients to be resotored. To improve color-image denoising in the redundant Haar wavelet domain, this paper studies marginal prior distributions of wavelet coefficients of natural clean color images, and addresses the propriety of the three color shrinkage schemes.
In this paper, we propose the reduction method of blocking artifact in the JPEG coded image using Pinned Sine Transform (PST). As a block noise reduction method, the Zero-Masking method which is adjusting some speci c DCT coefficients to 0 was proposed in the past. It has a problem that the block boundary fails by this noise reduction method. Our proposed method maintains information on the block boundary by PST, and prevents the block boundary from failing by its noise reduction. By using PST, the blocking artifact can be reduced without generating some new boundary between processed blocks even if the coefficient of the frequency domain is changed. As a reduction effect of blocking artifact, the score of SSIM of the proposed method has improved compared with the conventional Zero-Masking method and the subjectively image quality has also improved well.