Image analysis – Image compression or coding – Quantization
Reexamination Certificate
1998-07-10
2001-12-11
Tran, Phuoc (Department: 2721)
Image analysis
Image compression or coding
Quantization
C382S239000
Reexamination Certificate
active
06330369
ABSTRACT:
BACKGROUND
Sequences of digital images, obtained for example by digitizing motion pictures or television signals, commonly are compressed for storage on a computer readable and writeable medium for subsequent authoring and playback of multimedia programs using editing systems such as the Media Composer editing system from Avid Technology, Inc. A variety of compression techniques have been developed, including several standards. In some systems each image, i.e., a field or frame, in the sequence of digital images is compressed separately using still image compression techniques. An example of such a compression technique is known as “JPEG,” which is an acronym for 37 Joint Photographic Experts Group.” This group developed the ISO 10918-1 JPEG Draft International Standard, CCITT Recommendation T.81.
The draft standard is described in JPEG Still Image Data Compression Standard by William B.
Pennebaker and Joan L. Mitchell, New York: Van Nostrand Reinhold, 1993, and in “The JPEG Still Picture Compression Standard” by Gregory K. Wallace, Communications of the ACM, April 1991, pages 31-44.
Using JPEG and other forms of still image compression, the image is subdivided into blocks of picture elements (pixels). Each block is transformed from its color representation in the spatial domain to a color representation in a frequency domain, for example using a discrete cosine transform. The resulting matrix of frequency coefficients, one coefficient for each frequency, is quantized using a set of quantizers, one quantizer for each frequency, to provide a quantized value for each frequency. Each frequency coefficient is divided by the corresponding quantizer. The set of quantizers typically is referred to as a quantization table or quantization matrix. The quantized values are entropy-encoded. In the JPEG standard, entropy encoding is performed by run length encoding followed by Huffinan encoding. Arithmetic coding also may be used.
Adaptive quantization methods change the quantizers, either within an image on block boundaries, or from one image to the next. Quantizers may be modified either to change the data rate or to change the visual fidelity with which the image is reproduced. An increase in a quantizer value decreases the data rate but also loweres the fidelity of the quantized result. Quantizers may be adapted individually, or the table of quantizers may be scaled uniformly by a scale factor. The JPEG standard does not provide for adapting quantizers within an image. However when JPEG is applied to a sequence of digital images, the quantizers may be adapted from one image to the next. The total amount of data produced by compressing an image is compared to a target data rate, from which an adjustment to the quantization table is determined. The adjusted quantization table is applied to the next image in the sequence of digital images. Such a technique is described in U.S. Pat. Nos. 5,577,190 (Peters) and 5,355,450 (Garmon et al.).
SUMMARY
In a sequence of digital images a simple image may be followed by a more complex image. When such a sequence is compressed using adaptive quantization, a complex image following a simple image may cause a large change in data rate to occur. This large change occurs because the degree of quantization of the simple image is lowered to increase the data rate to the target data rate, which in turn increases the data rate when a complex image is encountered. The increased data rate can negatively impact performance.
In some applications, many sequences of digital images are combined together, using a technique called compositing. Two sequences are composited at a time to produce what is called a layer of the composite. Each layer and the final multi layer composite typically are compressed and stored. If each layer is compressed using a different quantization table, which is likely to occur when using adaptive quantization, image quality of the composite may degrade significantly.
By limiting the extent to which quantization is changed to increase the amount of compressed data, these problems may be reduced. In particular, when a more complex image occurs after a simple image, the quantization used to compress the complex image will not cause as large of a change in the total amount of compressed data. Recovery from such a change also may occur more quickly. In one embodiment, quantization tables used for an image may be adjusted using a scaling factor, and a limit on the scaling factor may be established.
When rendering multi layer composites, this limit is such that recompression of previously compressed data does not result in additional loss of information. As a result, degradation of image quality in each layer of the composite is avoided. In one embodiment, where quantization tables for an image are adjusted using a scaling factor, a limit on the scaling factor is established such that the same quantization table is used for to compress each layer produced of the composite.
Accordingly, in one aspect a method or apparatus for compressing a sequence of digital images performs compression on an image in the sequence of digital images according to a compression parameter to provide an amount of compressed data. When the amount of compressed data is greater than a target data rate, the compression parameter is adjusted such that a smaller amount of compressed data is obtained. When the amount of compressed data is less than the target data rate, the compression parameter is adjusted such that a larger amount of data is obtained. The adjustment to the compression parameter is limited according to a limit factor. In one embodiment, the compression parameter includes a quantization table and a scale factor. Performing compression includes transforming blocks of the image into matrices of values in a frequency domain and quantizing the matrices of values using a quantization table scaled by a scale factor. Adjusting the compression parameter thus involves changing the scale factor subject to the limit factor. In one embodiment, performing compression involves compressing an image in the sequence of digital images, and, after adjusting the compression parameter, a next image in the sequence of digital images is compressed using the adjusted compression parameter. In another embodiment, performing compression involves compressing a block of an image in the sequence of digital images and, after adjusting the compression parameter, a next block in the image is compressed using the adjusted compression parameter.
In another aspect, an apparatus for compressing a sequence of digital images includes and encoder and a controller. The encoder has a first input for receiving uncompressed image data, a second input for receiving a signal indicating a compression parameter for compressing the uncompressed image data, and an output for providing an amount of compressed data according to the compression parameter. The controller has a first input for receiving a signal indicating the amount of compressed data, a second input for receiving a signal indicating a target amount of data and a third input for receiving a limit factor, and an output providing a compression parameter to the encoder according to the difference between the target amount of data and the amount of compressed data and the limit factor. In one embodiment, the encoder compresses the image using a quantization table scaled by a scale factor. In this embodiment, the controller further comprises a circuit having a first input for receiving a signal indicating the amount of compressed data and a signal indicating the target amount of compressed data and an output for providing an updated scale factor. A comparator has a first input for receiving the limit factor and a second input for receiving the updated scale factor, and an output for providing a signal indicating a selected scale factor according to a comparison of the updated scale factor to the limit factor.
In another aspect, a method and apparatus for performing multi layer compositing of sequences of digital images involves generating a first co
Cornog Katherine H.
Hoag David F.
Avid Technology Inc.
Gordon Peter J.
Tran Phuoc
LandOfFree
Method and apparatus for limiting data rate and image... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Method and apparatus for limiting data rate and image..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method and apparatus for limiting data rate and image... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2600611