Image analysis – Histogram processing – For setting a threshold
Patent
1993-05-13
1995-03-07
Mancuso, Joseph
Image analysis
Histogram processing
For setting a threshold
348405, 348420, 258433, C06K 936
Patent
active
053965671
DESCRIPTION:
BRIEF SUMMARY
BACKGROUND OF THE INVENTION
A conventional mode of procedure in data reduction for the purpose of the transmission of digital images or digital sequences is the subdivision of image regions into smaller regions, so-called blocks (FIG. 2). These blocks are typically 8--8 or 16--16 pixels (picture elements) in size (FIG. 3). Examples of such block-related processes for image coding are the so-called transformation processes, of which the most frequently applied representative is the so-called discrete cosine transformation (D.C.T.). Signal flow diagrams for coding processes of this type for individual images or image sequences are represented in FIGS. 1a and 1b.
The purpose of coding is data reduction, by means of which digital transmission of images or image sequences only becomes possible in many instances. A known principle in this regard is the so-called redundancy reduction, in which statistic characteristics of the image signals are utilized for data reduction. The sole use of redundancy reduction guarantees the complete reconstructability of the image information from the coded data. However, with the aid of redundancy reduction alone it is normally possible to achieve only a data compression by factors of 2 to 3. In most applications of image coding and image sequence coding, however, there is a need for higher compression factors, for example at least a compression factor of 7 in the case of high definition television (HDTV). In order to achieve such compression factors and higher, such as in the case of videophone, it is also necessary to use methods of irrelevance reduction. However, in this case the original signal information is partially lost and image transmission errors are produced. The aim is, however, for these image errors not to be visible to a viewer or to disturb as little as possible.
In a transformation coding, redundancy reduction and irrelevancy reduction are generally applied jointly to the signal values in the transformation region of an image block (A. N. Netravali, W. G. Haskell: "Digital pictures, representation and compression" Plenum Press, New York, 1988). In DCT, these are the so-called DCT coefficients. Typical modes of procedure in irrelevance reduction are, on the one hand, simply to leave out entire subsets of DCT coefficients, and on the other hand to quantize the remaining, significant coefficients more coarsely, that is to say to represent them in a coarser raster of values than originally assigned to them. A raster of values can be characterized by a parameter Q for its fineness. The larger the Q, the coarser the raster. In a so-called linear quantization, this parameter is simply the quantization step width, that is to say the uniform interval between two adjacent raster values. Different types of quantization of DCT coefficients are known. In some processes, the individual coefficients are quantized differently in terms of fineness. The different finenesses are derived in this case from assumptions concerning the visual significance of the coefficients. Such a process is described, for example, in D. McLaren, D. T. Nguyen: "Video bitrate reduction through psychovisual impression", Proc. 1990 Video Communications Workshop, Melbourne, 9 to 11 Jul. 1990. In another process, all coefficients of equal fineness are quantized, as is described, for example, in W. Tengler, W. Ja.beta.: "Interlaced and Progressive HDTV-Coding, a Comparison for 140 Mbit/s- Transmission" Proceedings 1990 Austr. Video Communication Workshop Melbourne, 9 -11 Jul. 1990.
Different types of quantization generate different types of image errors. In the case of a transformation such as DCT, however, these image errors are generally distributed in each case over an entire block in the image region, irrespective of the quantization scheme employed. Whether these errors are either not visible at all or are more or less disturbing visually, depends on the original image content, on the one hand, and on the mean amplitude of the errors in a block, on the other hand. This mean error amplitude is directl
REFERENCES:
patent: 4677479 (1987-06-01), Hatori et al.
patent: 5021891 (1991-06-01), Lee
patent: 5101280 (1992-03-01), Moronega et al.
patent: 5241395 (1993-08-01), Chen
Gimlett, Use of "Activity" Classes in Adaptive Transform Image Coding, IEEE Transactions on Communications Technology, Jul. 1975, pp. 785-786.
"Video Bit-rate Reduction Through Psychovisual Compression", by David L. McLaren et al, Australian Video Communications Workshop, Melbourne, Jul. 9-11, 1990, pp. 247-254.
"Interlaced and Progressive HDTV-Coding--A Comparison for 140 MBIT/S--Transmission", by Walter Tengler, Australian Video Communications Workshop, Melbourne, Jul. 9-11, 1990, pp. 7-18.
Mancuso Joseph
Siemens Aktiengesellschaft
LandOfFree
Process for adaptive quantization for the purpose of data reduct does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Process for adaptive quantization for the purpose of data reduct, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Process for adaptive quantization for the purpose of data reduct will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-1412908