This article describes an application of differential prediction coding to data compression of digital image with gray levels using bilinear surface patches. The region on which a image is defined is subdevided into nonuniform blocks, and the luminance values in a block are approximated by bilinear surface patch constracted by the luminance values at the four corners of the block. We propose four methods to predict luminance value at the block's corner. Our coding method provides higher quality at low bitrate than JPEG method.
This paper proposes a new method to detect scene changes from MPEG encoded video data without decoding. Since the method uses much information from MPEG data, it can detect not only usual scene changes between frame, but also scene changes between fields and dissolve changes except effects of motion and flash light. This paper also proposes MPEG Video Browser using proposed Scene Change Detection method for rapid browsing of video data, and indicate the efficiency of the method.
Abstract Digitally coded television sequences tend to produce scene-dependent and temporally varying distortion at low bit rates. It is not unusual for a large distortion to appear only every 1O or 20 min., while the quality is otherwise good. Long sequences are necessary to appreciate the influence from quality variations. We propose a new method to evaluate the temporal variations of picture quality. It is, consequently, found that temporally varying distortions can be expressed by means of the worst and best scores per assessor for each sequence to some extent. Thus, our method can improve the low reliability of the SSCQE method proposed at the ITU-R meeting.
ITU-T Recommendation H.223 is an efficient way to multiplex various kinds of multimedia information, such as audio and video, and is widely used for a TV phone. H.223,however, is very vulnerable to errors and its extension for a mobile TV phone has been studied. In this paper, we analyze throughput performances of multiplexing scheme adopting PN flags instead of HDLC flags and derive the optimum PN length depending on the error conditions. Furthermore we propose a new schome to avoid flag emulations caused by the use of PN flags.
This paper surveys recent developments in image data compression by using variable size blocks. Following topics are described briefly : Data compression of images with gray level using triangular plane patches : Hierarchical picture description using triangular blocks and its application to contour tracing : JPEG based on 8x8 DCT ; Variable block size DCT ; Image coding by non-uniform block decomposition ; Block truncation coding ; Fractal image coding using variable shapes ; Adaptive image coding based on variable block shapes ; and Hierarchical coding of binary images. An efficient image coding is attained by the use of variable size and shape blocks.
This report proposes a hierarchical segmentation using the HSI color model for image coding of electric power facilities, and describes the extraction result of this method. The HSI color model is used because electric power facilities have no coloring. The hierarchical segmentation which is based on the k-mean clustering, is able to solve the decision problem of the number of clusters and speed up the processing owing to using stop conditions.
We developed the positioning method of vibrating lines from video stream which was captured by ITV-based monitoring system. We applied our method to video data which we recorded in order to observe how power line had vibrated. We compared frequency power spectrum of vibrating power line among our method, tension sensor which was set up in power line system and wind force datum based simulation of power line vibration. When the frequency characteristic of power line vibrating was less than 1.0 Hz (vibrating line's frequency is mainly less than 1.0 Hz), it gave good agreement with our method.