Difference Adaptive Block Truncation Coding (DABTC) is a high efficient coding scheme and produces good quality for still images. At low bit-rates, however, block noise often appear in certain DABTC coded images. In this report. Floating Difference Adaptive Block Truncation Coding (F-DABTC) is proposed. The F-DABTC divides a picture into NXN sized blocks, and each line of blocks is floated randomly in the direction of the horizon. Because of floating, the visibility of block noise in certain DABTC coded images at low bit-rates can be reduced.
Image partitions are used to find corresponding points between a pair of stereo images. The stereo images are first expressed by feature lines, and then these lines are partitioned into blocks on the basis of proposed degree of adjacency. These blocks are used as the units of correspondence. In correspondence, a coarse-fine procedure is taken into consideration, and the size of blocks are controlled at each correspondence stage. Correspondence between blocks of two images is done by using fuzzy operations of block properties, such as number of lines, directions, length. Experiment results show the effectivenesss of our approach.
Formerly we proposed the quantizer-selective DPCM coding scheme (QS-DPCM) for image signals. In this scheme, as the mean square error of each line is kept almost constant by selecting quantizers adequately, the average coding rate varies according to images. Thereupon, we propose a new method to equalize the average coding rate with any specified transmission rate. First, variances of prediction errors and the coding rate of each line are estimated on information of the several preceding lines. And then, quantizers are adaptively selected using these estimated parameters and also buffer memory occupancy. Computer simulation indicates that SNR of this scheme is 0.5〜1 dB lower than that of QS-DPCM and moreover requisite buffer memory capacity is about 2% of that in QS-DPCM.
A versatile desk top color scanner has been developed, using high rezolution contact color image sensor. Not only reflective but transparent originals can be read. Other feature : stop and re-start function, A3 size scanning area, high functional image processing, and three kinds of interface (SCSI is optional).
Authors have developed a CCD contact image sensor unit with high resolution (400 dots per inch) and high scanning speed (0.5 millisecond per line). This sensor consists of 12 CCD chips, all of which are arranged in a line, with uniform sensing pixel pitch. An allowable limit of pixel pitch error at the jointed portion has been determined by evaluation through questionnaires. As a result, the limit pitch is about two times as long as standard pitch. This paper describes as follows. (1) the relation between distortion of output image and pixel pitch error (2) the determination of allowable limit of pixel pitch error (3) the structure and specifications of the sensor unit
A ZrN thin-film resistor is proposed as a promising heating resistor for a thermal printing head used in high-resolution high-speed printing. The ZrN film is prepared by nonreactive sputtering with a hot-pressed ZrN target. Practical films, which have a small resistance change with temperature as well as during aging tests, are provided with resistivities of 0.5 to 5 mohm・cm. B7 size and 16-dot/mm resolution thermal printing heads, installed with driver LSI's, are fabricated to examine the applicability for high-speed printing. The results of performance tests provide the possibility that the heads are able to operate with speed below 1 msec/line.
NHK Laboratories are developing the facsimile broadcasting system for the data channel of DBS. The data channel of DBS which is going to be used in the packet format, is expected as the transmission line of many data broadcasting systems. The facsimile programs has the normal programs and the supplementary programs which are printed on thermal papers. The authors has developed the intermitted transmission method for the facsimile system and make clear the characteristics of the facsimile pictures in the data channel of DBS. One facsimile page which is consisted of half torn pictures and normal size characters, is printed out less than 30 Sec. by this method.
This paper describes characteristics of video coding data and interval of packetized video coding data stream. This video coding is based on motion compensated adaptive intra-interframe predictive coding argorithm which maintains SN ratio constant by changing quantization step. This paper also presents a generating process for interval of packetized video data stream.