Pulse or digital communications – Bandwidth reduction or expansion – Television or motion video signal
Reexamination Certificate
1998-11-30
2001-02-27
Kelley, Chris S. (Department: 2713)
Pulse or digital communications
Bandwidth reduction or expansion
Television or motion video signal
C348S424200
Reexamination Certificate
active
06195394
ABSTRACT:
BACKGROUND
1. Field of the Invention
This invention relates to (1) pre-processing techniques for reducing the image spatial bandwidth of original digital motion pictures, to thereby reduce the overall encoded compressed data required to be forwarded to a decoder for decompressing the encoded compressed data, together with (2) post-processing techniques for synthetically extending the image spatial bandwidth of the decompressed digital motion pictures from the decoder.
2. Description of the Prior Art
As known, there are many popular still image and motion image encoding schemes presently in use, such as MPEG-1, 2, 4, & 7, JPEG, CD-i, H.261, and M-JPEG. The advantage of these algorithms is that they all are agreed upon worldwide standards, offer various compression ratios for various applications, and do their best to preserve image quality. Although all these schemes are considered very technologically advanced, they still introduce some level of image artifacts, which are often above visible threshold.
In particular, MPEG (Motion Picture Experts Group) encoding for motion video accomplishes image data compression by encoding both the motion and spatial redundancies in moving images in a cooperative way. The amount of data necessary to represent a series of image frames depends greatly on the statistics of the image sequence, and is not fixed by the input image resolution, the level of quantization of the image samples, or the number of frames per second. Therefore, the actual output data rate represents an average number of bits per image computed and based upon a specific sample set of image frames. More specifically, transmission channel bandwidth is a statistical function based on the averaged characteristics of the following three MPEG encoding components of a sequence or group of pictures (GOP):
(1) a motion-encoding component of MPEG, which is based on comparing ‘macro blocks’ in successive video frames and determining where there are redundancies (wherein a macro block typically comprising a 16×16 block of pixels) If they were the same, similar macro blocks in successive frames would not be re-coded, but instead, repeated and/or predicted based on motion estimation or MCP (block-based motion compensated prediction);
(2) a spatial-encoding component of MPEG, which is used to reduce redundancies by comparing neighboring macro blocks using the discrete cosine transform (DCT) to track changes in luminance and chrominance. Typically, picture data is block coded using a two-dimensional 8×8 DCT. The sixty three coefficients are mapped into a zigzag pattern, quantized, run-length coded, and Huffman coded; and
(3) a spatial-frame-redundancy component of MPEG, which is used to reduce data by one-dimensional differential pulse-code modulation (DPCM) coding of the zero-frequency or DC coefficients of frames, then quantization, and finally entropy coding. The amount of motion and high frequency information largely determine the bulk of data required to code a series of frames.
At the current level of technology, most hybrid types of “entropy-quantization” algorithms generally perform transparently at compression ratios of 10:1 or less. Unfortunately, many applications require much higher compression ratios, where under these conditions numerous artifacts may become visible. These artifacts often include Gibbs phenomenon, blockiness, posterization, checker boarding, and color bleeding.
Regardless of whether it is transmission channel bandwidth, the number of gigabits found on a DVD CD-ROM, or computer disk transfer rate, there is never more bandwidth for free. Therefore, normally, the image transmission bandwidth is the limiting factor in overall image quality because of the substantially large cost of providing greater image transmission. For this reason, in MPEG and like systems, a compromise is usually made between how ‘quantized’ the encoded image pixels are and how much motion change needs to be coded. This trade is a balancing of image statistics to keep the overall average bit rate lower than the channel bandwidth, so that an output image can always be reconstructed and the viewer maintains a picture on the screen.
There are many ways used by the prior art to reduce MPEG based encoded data rates. Typically, all these ways strive to remove all data redundancies and, assuming the algorithm employed is sound and the encoder employed provides a faithful implementation of the employed algorithm, there is little else to gain in coding efficiency. For example, a first prior-art solution to reduce overall channel bandwidth is to reduce the number of input samples that the encoder needs to process by starting with a smaller size input image, as described in “What is MPEG?” by Mark Adler, dated Oct. 19, 1992, which is published in MPEG-FAQ 4.1. A smaller input image containing fewer pixels would naturally take fewer bits to encode. A second prior-art solution to reduce overall channel bandwidth is to pre-filter the input image by some fixed amount, as described in “What is MPEG-2?” by Chad Fogg, dated May 11, 1995, which is published in MPEG-FAQ 4.1. The image would then contain less high frequency, and, therefore, would command a lower Nyquist sampling rate. Thus, the encoder could theoretically represent the image with fewer bits The pre-filtered image has less high frequency information, and the encoder would see fewer high frequency edges and also likely detect less motion. However, viewing a smaller size image (as suggested by the first prior-art solution) or a more blurred image (as suggested by the second prior-art solution) would not be considered an improvement in system image quality. So these first and second prior-art solutions are not optimal.
There are many observed cases in current MPEG applications where pixelization from over quantizing in high motion sequences already occurs and produces visible artifacts. This occurs because the encoder is economized and is operating near the limit of its abilities, and when occasionally it cannot compress enough the limit of the transmission channel is exceeded.
At this point, the question is what else might be done to either further reduce the coded data rate and/or how can the level of perceived image artifacts be reduced? In this regard, reference is made to the teachings of each of our two earlier U.S. Pat. No. 5,355,328 (which issued Oct. 11, 1994 and is entitled “Resampling Apparatus Suitable For Resizing a Video Image”) and U.S. Pat. No. 5,483,474 (which issued Jan. 9, 1996 and is entitled “D-Dimensional Fractional Bandwidth Signal Processing Apparatus”), and to the teaching of our copending patent application Ser. No. 09/112539 (which was filed Jul. 9, 1998, is entitled “Processing Apparatus for Synthetically Extending the Bandwidth of a Spatially-Sampled Video Image” and is assigned to the same assignee as the present invention). The respective teachings of each of these two patents and this patent application are incorporated herein by reference.
SUMMARY OF THE INVENTION
The present invention is directed to an improvement in a compressed data system for communicating high spatial-frequency content of original frames of a digital motion picture over a limited-bandwidth transmission channel which, while preventing the bandwidth of the compressed data from ever exceeding the limited bandwidth of the transmission channel, still permits these communicated frames, after being received and decompressed, to substantially retain their high spatial-frequency content. This is achieved by employing a pre-processor that includes first digitally-controlled means, responsive to a digital control value of a first digital control signal being applied as a control input thereto which is indicative of the bandwidth of the compressed content exceeding the limited bandwidth of the transmission channel, for reducing the spatial bandwidth of the content of each of original frame by a selected amount which is sufficient to maintain the bandwidth of the compressed content within the limited bandwidth of said transmiss
Arbeiter James Henry
Bessler Roger Frank
Kelley Chris S.
North Shore Laboratories, Inc.
Seligsohn George J.
Vo Tung
LandOfFree
Processing apparatus for use in reducing visible artifacts... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Processing apparatus for use in reducing visible artifacts..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Processing apparatus for use in reducing visible artifacts... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2604608