Ryoichi Kawada

Objective measurement scheme for perceived picture quality degradation caused by MPEG encoding without any reference pictures

The rapid progress in digital transmission technology has spurred demand for developing a technology that supports monitoring of video transmissions. Automating the picture quality assessment process is in particular demand because it currently depends on subjective assessments by human operators and places a heavy burden on them. The authors therefore propose an objective picture quality measurement method that works without reference pictures. In this proposed method, invisible markers are embedded into original pictures by use of the spread spectrum data hiding method. Since the markers are widely spread over the frequency domain, the degradation caused by MPEG compression can be estimated by detecting the extent of marker degradation. Our method is usable regardless of the kind of picture, bitrate, and number of stages of tandem codec connections. The degradation in picture quality caused by the embedded markers is quite small and not perceivable by the human eye. The proposed method is therefore applicable to a wide range of visual transmission services.

Effective rate control method for minimizing temporal fluctuations in picture quality applicable for MPEG-4 AVC/H.264 encoding

Appropriate rate control plays a very important role in encoding motion pictures under the constant bit-rate. One of the requirements for rate control is minimizing temporal fluctuations in picture quality, which may cause flicker artifacts. To satisfy the requirement, some methods for MPEG-1 and MPEG-2 have been proposed. However, these methods cannot be applied to the H.264 encoding algorithm based on R-D optimization since they require the DCT coefficients and the accurate output bit amount for each quantization parameter of a current picture in advance. To overcome the problem, we propose a rate control method to satisfy the requirement in the H.264 encoding algorithm by introducing approximations to estimate the bit amount and the distortion.

An efficient parallel video transmission system including codecs with functions of failure detection and coding noise evaluation

For contribution and primary distribution of television, i.e., television transmission to broadcasting stations, parallel transmission links that include codecs are often set up for reliability. However, this configuration still faces the following issues: 1) during normal periods, the reserve link is totally wasted and 2) when a failure occurs, the disrupted signal continues to be transmitted until the system switches to the reserve link. In addition, in digital video transmission, automatic assessment of received video quality is an important issue. To address these issues, the authors developed a processing system for parallelly transmitted video signals, which mainly consists of the following three elements: 1) when both links are normal, outputting the average signal of both links reduces compression coding noise and improves picture quality compared to only one link (efficiency); 2) comparison of the pictures from the two links enables failure detection in either link, to switch to the normal link (support of failure detection); and 3) comparison of the two normally decoded pictures enables evaluation of picture quality (automatic coding noise evaluation). This unique system improves the reliability and operational efficiency of conventional parallel video transmission links.

Fast and robust content-based copy detection based on quadrant of luminance centroid and adaptive feature comparison

This paper proposes a fast and robust content-based copy detection scheme. Our proposal consists of a new compact feature, efficient keyframe selection and adaptive mask-based feature comparison. Firstly, a block-level luminance centroid is binarized into a 32-bit quadrant feature for fast and robust feature comparison. Subsequently, a new keyframe selection method is adopted to enhance pairwise independence between unrelated video segments in addition to choosing stable keyframes. Finally, a block-level mask-based feature comparison method is introduced to compare only stable features. Experimental results show our scheme improves recall by 0.1 at the same precision 0.9 and the processing speed in feature comparison of the proposed scheme is about twice as fast as that of conventional schemes.

Prefilter Control Scheme for Low bitrate TV Distribution

In IP-based TV distribution, coding degradation is sometimes evident in critical scenes because the bit rate for compression is rather low. Prefiltering is an effective countermeasure since it replaces the coding noise with the degradation more difficult to detect visually, though it has the drawback that excessive smoothing might occur. This paper proposes a scene-adaptive method to control a prefilter separate from the encoder. By calculating block-wise motion-compensated predictive error variances and correlation coefficients, it estimates the coding noise as well as the potential improvement by prefiltering each frame, realizing a control scheme which performs prefiltering only when effective

Development of picture quality monitoring system for IPTV service based on the reduced reference framework

The authors developed a software-based realtime IPTV monitoring system based on Reduced Reference framework, and evaluated the proposed system. One of the quality issues of the IPTV service is the picture quality degradation caused by packet loss. The proposed system precisely estimates the PSNR of the corrupted received picture by extracting and comparing image features from transmission and receiver side. Computer simulations show that PSNR estimation with a 0.945 correlation coefficient at a data channel bitrate of 36kbps is possible using the proposed system.

Effective interpolation for free viewpoint images using multilayered dynamic background buffers

Presenting images from a free viewpoint is a promising interactive video application. Though reference images from one viewpoint and their depth maps are often used to render free viewpoint images, picture quality degradation may occur because of lack of information in background regions that are occluded by foreground regions. In this paper an interpolation method for free viewpoint images using multilayered dynamic background buffers is proposed. In the proposed method, the buffers, updated using the reference images divided in each frame, are used to store the background regions and the output images are interpolated by the background buffers. Since the background buffers are created and updated using only the reference images and their depth maps, additional information on the background buffers is not required for the interpolation. The effectiveness of the proposed method was evaluated by several simulation experiments.

Automatic image selection scheme utilizing comments for insertion of images into weblogs

This paper presents a scheme which utilizes comments given to images on an image-sharing site in order to obtain an appropriate image for insertion into poem-like weblogs (blogs) as a way to represent their atmosphere (impression). The result shows that utilizing comments is effective. To achieve this purpose, there are two issues: how impression words are extracted from blogs and how images representing the impression words are obtained. Assuming that it is important to obtain images representing the impression words, this paper focuses on only the latter issue. We hypothesize that comments and tags extracted from an image-sharing site can be adequate for obtaining images corresponding to impression words at low cost. In particular, utilizing comments can be more appropriate for the image search with impression words than utilizing tags because the impression words are often used in comments. Therefore, we propose a scheme which utilizes comments to obtain appropriate images. In order to investigate the effectiveness of utilizing comments, conformance between impression words and the images was evaluated. The rating for conformance is 3.5 on a scale of 1 to 5 when utilizing comments, which is 0.6 higher than when utilizing tags.

Contrast-enhancing and deterministic tone mapping method in natural image hiding scheme using halftone images

This paper presents a novel tone mapping method for a natural image hiding scheme using halftone images, where a natural image can be visually decoded by overlaying two natural halftone images. In this scheme, there is a tradeoff between noise and contrast, and a generally applicable setting of tone mapping before halftoning is required. However, in a conventional method, the coefficients for the tone mapping cannot be found automatically. It is difficult even for experts to manipulate the tradeoff with the coefficients. To solve this problem, we propose a deterministic tone mapping method that can intuitively control the tradeoff. To realize this, we introduce a noise metric which can be understood intuitively instead of affine parameters used in the mapping system. To maximize the dynamic range at a given noise level and to make tone mapping deterministic, we clarify the condition and introduce the general functions to obtain coefficients from the geometric constraint of the tone mapping region. The proposed method enables any user to generate images with the highest possible contrast at a given noise level deterministically from any natural images such as own photos just by setting the noise level. Experimental results show the validity of the proposed noise metric and also show that the generally applicable tradeoff point through various images that are used as a guideline to set the noise level. By using the tradeoff point, the average dynamic range is expanded by a factor of 1.4 compared to the noiseless case.

4-1 Human Motion Tracking in a Single Camera with a Graph Structure in Latent Spaces

This paper presents a novel approach to track human motion with a monocular video based on particle filter. To deal with the high dimensional freedoms in human pose, a low dimensional latent space, e.g. Gaussian Process Dynamical Model (GPDM), is a typical solution. In this paper, we extend the GPDM model into a graph structure to better express the diverse dynamics in human motion, where multiple latent spaces are constructed and connected with each other.