Electrical computers and digital processing systems: memory – Storage accessing and control – Hierarchical memories
Reexamination Certificate
1996-12-02
2001-01-09
Peikari, B. James (Department: 2752)
Electrical computers and digital processing systems: memory
Storage accessing and control
Hierarchical memories
C711S125000, C712S004000, C710S066000, C710S068000, C710S120000
Reexamination Certificate
active
06173366
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
This invention relates to digital signal processing (DSP), and more particularly to an extension unit added to a microprocessor for high speed multimedia applications. The extension unit includes an operand routing unit which aligns multiple operands upon an arithmetic logic unit (ALU) in response to specific multimedia-type instructions. Proper ordered arrangement of operands at the ALU enhances the throughput of many image compression algorithms which rely upon repetitive, sequential operations.
2. Description of the Relevant Art
It is well known that conventional computers communicate information primarily through a graphical user interface (GUI). The GUI involves manipulation of complex graphical images, as either still graphic images or full motion video. Current software has spawned numerous multimedia applications which require administering still images or video via the GUI.
Processing still images or video consumes prodigious amounts of storage space within the computer. For example, a 256 color VGA screen image can entail numerous rows and pixels, each consuming a single byte of store. For example, a partial screen containing 200 rows of 320 pixels consumes a minimum of 64K bytes of storage. Real time processing of still images (and especially video) thereby requires that the amount of data be reduced. The task of reducing the amount of data necessary to store or transmit one or more digital images is often referred to as “image compression”.
Image compression can be classified as either lossy or lossless. If the reconstructed image is not identical to the original image, the compression is said to be lossy. Lossy compression is used where the reconstructed image, while not identical to the original image, nonetheless conveys the essential features of the image. Minor changes may not be perceptible to a human observer, or may not be objectionable for a particular application. Lossy compression can therefore reduce the amount of data relative to lossless compression but without perceptible defects.
FIG. 1
illustrates a conventional lossy image compression system
10
. System
10
is shown applicable to image (i.e., still image or full motion image) compression and decompression. An original image is compressed by an image encoder
12
, and the encoded output may be further processed in block
14
using, for example, error correction, encryption, multiplexing, etc. The compressed image can be stored or sent through a communications channel. If forwarded through a communications channel, the compressed data is modulated upon a carrier signal by modulator
16
. The data-modulated carrier signal is then forwarded to a decoder via channel
18
. If the data is transmitted and requires demodulation, block
20
is used to extract the compressed image which can then be further processed as needed by block
22
. Block
22
is used to perform, for example, decryption, demultiplexing, etc. Decoder
24
receives the compressed image having redundant or irrelevant data removed, and thereafter produces a reconstructed image perceptibly similar to the original image.
FIG. 2
illustrates, in further detail, an image encoder
12
used for compressing an image as either a still image or a sequence of images (i.e., full motion video). Upon receiving the image in either RGB or YCrCb format, encoder
12
encodes certain “frames” of a plurality of frames within the sequence of motion images or still images. Frames within a video sequence can be compressed using numerous compression standards, a popular one being the Moving Pictures Experts Group (MPEG) standard. MPEG compression involves discerning intracoded frames from non-intracoded frames. An intracoded frame, often called I-frame, is compressed relative to itself, while a non-intracoded frame, often called P-frames and B-frames, are encoded by exploiting temporal redundancy as well as spatial redundancy to reduce the number of bits required for encoding.
Encoding and decoding video presents many challenges to realizing an efficient MPEG compression standard. The intracoded frames are stored, generally in a moderately compressed format. Successive non-intracoded frames are compared with the intracoded frames and the differences are stored. Periodically, such as when a new scene is displayed, a new intracoded frame is stored, and subsequent comparisons begin from this new reference point.
Video compression standards such as MPEG, DVI and Indeo, all use the intracoded frame technique. Many compression standards such as MPEG treat various frames within the frame sequence as a still image and apply still image compression to those frames. A popular still image compression standard is Joint Photographic Experts Group (JPEG). Encoder
12
illustrates numerous blocks used in MPEG video compression, of which a portion of those blocks are pertinent to, e.g., JPEG. The JPEG portion of encoder
12
is shown within dashed area
26
. Functional blocks within dashed area
26
serve to compress pixel data within blocks of each macro block arising from the original frame or image. The compressed digital data is then forwarded into an embedded decoder
30
. Embedded decoder
30
is used in a feedback arrangement, wherein the output of decoder
30
is subtracted from the original frame. Subtraction is shown at block
32
, and the output from functional blocks
26
is shown fed into a buffer
34
for subsequent output as compressed intracoded and non-intracoded frames.
In order to avoid having to store or transmit large amounts of information on each pixel within each frame, MPEG reduces the data to that which is pertinent only to intracoded and non-intracoded frames. As seen in the feedback arrangement of
FIG. 2
, data manipulation must be performed as rapidly as possible on each macro block or frame, preferably in real time. Substantial data reduction (lossy compression) is needed on frames of interest and generally occurs in JPEG blocks
26
and, more specifically, during quantization.
JPEG generally employs three stages of compression. A first stage utilizes a discrete cosine transform (DCT) function
36
. DCT is a class of mathematical operations which take a signal and transform it from one type of representation to another. Specifically, DCT converts an array of numbers, which represent signal amplitudes at various points in time and space, into another array of numbers, each of which represent the amplitude of a certain frequency component from the original signal. The resulting array of numbers contains the same number of values as the original array. Using a JPEG format, DCT transform is performed on a block of 8×8 picture elements (or “pels”) taken from an original image.
Output from DCT
36
is fed to a quantizer
38
. Quantization
38
involves the lossy stage of data compression by reducing the number of bits needed to store an integer value of lessened precision. A quantization matrix, chosen by a code word, reduces the matrix values output from DCT to the indices for the code words. Upon decode, the images are reconstructed using a table look-up procedure, given the code word selected by the quantization algorithm. The International Standards Organization (ISO) maintain the quantatization code words used by implementers of JPEG code. The quantization matrix can be coded in block
40
using several methods. For example, the quantized images of each frame can be arranged in a zig-zag sequence. The zig-zag sequence is then coded using run-length encoding (RLE) followed by entropy coding (which includes the popular Huffman code).
Code output from block
40
is a variable length code which generally represents smaller decimal numbers, and can be represented with corresponding smaller number of bits depending upon the decimal value. An advantage of using smaller number variable length coding is carried forth within the intracoded and non-intracoded sequence of frames, or more particularly within each macro block of a frame. Accordingly, MPEG involves JPEG-type compression on each sel
Favor John G.
Thayer John S.
Weber Frederick D.
Compaq Computer Corp.
Conley & Rose & Tayon P.C.
Daffer Kevin L.
Kowert Robert C.
Peikari B. James
LandOfFree
Load and store instructions which perform unpacking and... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Load and store instructions which perform unpacking and..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Load and store instructions which perform unpacking and... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2516149