QUANTIZATION AND DIFFERENTIAL CODING OF ALPHA IMAGE DATA

• US 20090284651A1
• Filed: 05/19/2008
• Published: 11/19/2009
• Est. Priority Date: 05/19/2008
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Srinivasan, Sridhar
• Granted Patent: US 8,208,543 B2
• Time in Patent Office: Days
• US Class Current: 348/425.100
• CPC Class Codes: H04N 19/126 Details of normalisation or weighting functions, e.g. normalisation matrices or variable uniform quantisers
• CPC Class Codes: H04N 19/21 with binary alpha-plane coding for video objects, e.g. context-based arithmetic encoding [CAE]
• CPC Class Codes: H04N 19/50 using predictive coding H04N19/61 takes precedence

Abstract

Alpha images are efficiently encoded for inclusion in video bitstreams. During encoding, alpha image data is quantized as a first step before differential encoding. The quantized alpha image data is then differential encoded and represented in a modulo domain before entropy encoding to take advantage of certain distribution features of typical alpha image data. During decoding, a decoder performs differential decoding of encoded alpha image data before dequantization. During differential decoding, the data is converted from a modulo domain to a continuous domain. Dequantization is performed using a technique which results in reconstructed alpha image values which include 0 and maximal values within the acceptable alpha image data range so that the process preserves these values for reconstructed alpha images.

Claims

1. One or more computer-readable media containing computer-executable instructions which, when executed by a computer, cause the computer to perform a method for decoding an alpha image encoded in a bit stream, the method comprising:

entropy decoding encoded data from the bit stream to produce differential residual data for the image, the differential residual data encoded in a modulo domain;

differential decoding the residual data to produce quantized data for the alpha transparency image, the quantized data represented in a continuous domain; and

dequantizing the quantized data from the continuous domain to produce reconstructed data for the alpha image such that a quantized data value representing a full encoded alpha image value is dequantized to result in a full reconstructed alpha image value.

2. The method of claim 1, wherein differential decoding the residual data comprises

differential decoding the residual data in the modulo domain to reconstruct quantized data in the modulo domain;

and

performing a modulo operation on the reconstructed quantized data in the modulo domain to produce quantized data in the continuous domain.

3. A method for encoding an alpha transparency image in a video sequence, the method comprising:

representing data for the alpha transparency image in a modulo domain;

transmitting the data in the modulo domain in a video bit stream.

4. The method of claim 3, further comprising quantizing data for the alpha transparency image, and wherein representing data in the modulo domain comprises representing the quantized data in the modulo domain.

5. The method of claim 3, where data for the alpha transparency image is represented in the modulo domain without quantizing, and where alpha transparency data is losslessly encoded in the video sequence.

6. The method of claim 4, wherein representing the quantized data in the modulo domain comprises:

differential encoding the quantized data;

representing the differentially-encoded data into a modulo domain.

7. The method of claim 6, wherein the differentially-encoded data is represented in the modulo domain using a signed modulo operation.

8. The method of claim 6, wherein differential encoding for a quantized data value comprises using a prediction from a value to the left of the quantized value.

9. The method of claim 6, wherein differential encoding for a quantized data value comprises using a prediction from a value to the top of the quantized value.

10. The method of claim 6, wherein differential encoding for a quantized data value comprises using a prediction from values to the left of and to the top of the quantized value.

11. The method of claim 6, further comprising including one or more signals in the video sequence to indicate one or more types of differential encoding rules used for quantized data in the video sequence.

12. A method for decoding an alpha transparency image in a video sequence, the method comprising:

receiving a video bit stream containing alpha image data in a modulo domain;

transforming the modulo domain alpha image data into quantized data in a continuous domain; and

dequantizing the quantized data to produce reconstructed alpha transparency data.

13. The method of claim 12, further comprising entropy decoding encoded data received from the video bit stream.

14. The method of claim 12, wherein the modulo domain alpha image data comprises differential encoded residual data and wherein transforming the modulo domain alpha image data into quantized data comprises differential decoding the residual data.

15. The method of claim 14, wherein differential decoding the residual data comprises:

differential decoding the residual data to reconstruct quantized data in a modulo domain; and

performing a modulo operation on the reconstructed quantized data in the modulo domain to produce quantized data in a continuous domain.

16. The method of claim 15, wherein differential decoding residual data comprises one or more of:

decoding using a left prediction rule and decoding from a prediction from above rule.

17. The method of claim 15, wherein differential decoding residual data comprises one decoding using a prediction rule which relies on predictions from the left and the top.

18. The method of claim 16, wherein differential decoding rules are used according to codes received in the video bit stream.

19. The method of claim 15, wherein differential decoding is performed in parallel.

20. The method of claim 12, wherein dequantizing the quantized data comprises performing a dequantization operation on the quantized data which maps maximal and minimal quantized data values to maximal and minimal reconstructed alpha image data values, respectively.