High dynamic range data format conversions for digital media

• US 8,880,571 B2
• Filed: 05/05/2006
• Issued: 11/04/2014
• Est. Priority Date: 05/05/2006
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Srinivasan, Sridhar, Zhou, Zhi
• Primary Examiner(s): Ngo, Chuong D
• Assistant Examiner(s): Hughes, Kevin G
• Application Number:
  US11/418,627
• Publication Number:
  US 20070258641A1
• Time in Patent Office: 3,105 Days
• Field of Search: 708/204, 712/222, 382232-233
• US Class Current: 708/200
• CPC Class Codes: H03M 7/24 Conversion to or from floating-point codes
• CPC Class Codes: H04N 1/407 Control or modification of tonal gradation or of extreme levels, e.g. background level
• CPC Class Codes: H04N 1/6027 Correction or control of colour gradation or colour contrast H04N1/6058 takes precedence

Abstract

One or more continuous mappings are defined at a digital media encoder to convert input digital media data in a first high dynamic range format to a second format with a smaller dynamic range than the first format. The encoder converts the input digital media data to the second format with the smaller dynamic range using the continuous mapping and one or more conversion parameters relating to the continuous mapping. The encoder encodes the converted digital media data in a bitstream along with the conversion parameter(s). The conversion parameter(s) enable a digital media decoder to convert the converted digital media data back to the first high dynamic range format from the second format with the smaller dynamic range. Techniques for converting different input formats with different dynamic ranges are described.

Claims

1. A method for encoding digital media comprising:

by a computing device configured to implement the method, receiving digital media data in a first format, the digital media data in the first format including a plurality of still image sample values that collectively form a still image, each of the plurality of sample values of the digital media data in the first format comprising a sign bit, an exponent and a mantissa having a first bit length;

by the computing device configured to implement the method, converting the still image sample values from the first format to a second format, including, for each of the plurality of still image sample values, obtaining a scaled exponent by scaling the exponent for the sample value with a specified exponent bias value and adjusting the bit length of the mantissa to a second bit length different from the first bit length, the digital media data in the second format having a different dynamic range than the first format, each of the plurality of sample values of the digital media data in the second format comprising the scaled exponent scaled by the specified exponent bias value, the mantissa for the sample value, and no separate sign bit;

by the computing device configured to implement the method, encoding the converted still image sample values; and

by the computing device configured to implement the method, generating a bitstream comprising the encoded still image sample values, a first bitstream syntax element that indicates the exponent bias value for the encoded still image sample values, a second bitstream syntax element that indicates the second bit length of the mantissa, and a third bitstream syntax element that indicates the first format from among multiple possible source formats and that indicates the bit depth and format type using a single syntax element, wherein the first bitstream syntax element, the second bitstream syntax element, and the third bitstream syntax element are part of the image header layer of the bitstream.

2. The method of claim 1 wherein the second bit length is shorter than the first bit length.

3. The method of claim 1 wherein the second bit length is longer than the first bit length.

4. The method of claim 1 wherein the adjusting the bit length of the mantissa to a second bit length is based on an adjustable mantissa length conversion parameter.

5. The method of claim 4, wherein the exponent bias value is based on an adjustable exponent bias value conversion parameter.

6. The method of claim 4, wherein the digital media data in the first format is high dynamic range (HDR) image data.

7. The method of claim 1 further comprising, based on the sign bit, performing a 2'"'"'s complement operation as part of the converting.

8. The method of claim 1 further comprising decoding the encoded still image sample values at a digital media decoder.

9. The method of claim 1 wherein the first and second bitstream syntax elements are only included in the bitstream when the third syntax element indicates that the source image has a floating point format.

10. The method of claim 1, wherein the specified exponent bias value is specified by a user.

11. The method of claim 1, wherein the encoding the converted still image sample values further comprises (a) performing a scaling operation to reduce rounding errors during intermediate encoding operations by multiplying the still image sample values at a front end of the encoder;

(b) performing the intermediate operations on the scaled still image sample values; and

(c) dividing final values by an amount corresponding to the multiplying, and

wherein the generating the bitstream further comprises generating a fourth syntax element indicating whether scaling was performed.

12. A computer-readable storage medium storing computer-executable instructions for performing a method, the method comprising:

receiving digital media data in a first format, the digital media data in the first format including a plurality of still image sample values, each of the plurality of still image sample values of the digital media data in the first format comprising a sign bit, an exponent and a mantissa having a first bit length;

converting the still image sample values from the first format to a second format, including, for each of the plurality of still image sample values, obtaining a scaled exponent by scaling the exponent for the sample value with a specified exponent bias value and adjusting the bit length of the mantissa to a second bit length different from the first bit length, the digital media data in the second format having a different dynamic range than the first format, each of the plurality of sample values of the digital media data in the second format consisting of the scaled exponent scaled by the specified exponent bias value and the mantissa for the sample value;

encoding the converted still image sample values; and

generating a bitstream comprising the encoded still image sample values, a first bitstream syntax element that indicates the exponent bias value for the encoded still image sample values, a second bitstream syntax element that indicates the second bit length of the mantissa, and a third bitstream syntax element that indicates the first format from among multiple possible source formats and that indicates the bit depth and format type using a single syntax element, wherein the first bitstream syntax element, the second bitstream syntax element, and the third bitstream syntax element are part of the image header layer of the bitstream.

13. The computer-readable storage medium of claim 12, wherein the second bit length is shorter than the first bit length.

14. The computer-readable storage medium of claim 12, wherein the second bit length is longer than the first bit length.

15. The computer-readable storage medium of claim 12, wherein the adjusting the bit length of the mantissa to a second bit length is based on an adjustable mantissa length conversion parameter.

16. The computer-readable storage medium of claim 12, wherein the method further comprises, based on the sign bit, performing a 2'"'"'s complement operation prior to outputting encoded digital media data.

17. The computer-readable storage medium of claim 12, wherein the exponent bias value is based on an adjustable exponent bias value conversion

18. The computer-readable storage medium of claim 12, wherein the bitstream further includes a bitstream syntax element indicating a color format of a source image.

19. The computer-readable storage medium of claim 12 wherein the first and second bitstream syntax elements are only included in the bitstream when the third syntax element indicates that the source image has a floating point format.

20. The computer-readable storage medium of claim 12, wherein the specified exponent bias value is specified by a user.

21. The computer-readable storage medium of claim 12, wherein the digital media data in the first format is high dynamic range (HDR) image data.

22. The computer-readable storage medium of claim 12, wherein the encoding the converted still image sample values further comprises (a) performing a scaling operation to reduce rounding errors during intermediate encoding operations by multiplying the still image sample values at a front end of the encoder;

(b) performing the intermediate encoding operations on the scaled still image sample values; and

(c) dividing final values by an amount corresponding to the multiplying, and

wherein the generating the bitstream further comprises generating a fourth syntax element indicating whether scaling was performed.

23. A method for decoding digital media comprising:

by a computing device configured to implement the method, receiving a bitstream comprising encoded still image sample values, a first bitstream syntax element that indicates the exponent bias value for the encoded still image sample values, a second bitstream syntax element that indicates a first bit length of the mantissa, and a third bitstream syntax element that indicates a second format from among multiple possible formats and that indicates the bit depth and format type of the second format using a single syntax element, wherein the first bitstream syntax element, the second bitstream syntax element, and the third bitstream syntax element are part of the image header layer of the bitstream;

by the computing device configured to implement the method, decoding the encoded still image sample values into still image sample values in a first format; and

by the computing device configured to implement the method, converting the still image sample values in the first format into still image sample values in the second format using the first bitstream syntax element, the second bitstream syntax element, and the third bitstream syntax element,

each of the still image sample values in the first format comprising a scaled exponent scaled by the exponent bias value specified by the first bitstream syntax element, a mantissa having the first bit length specified by the second bitstream syntax element, and no separate sign bit,

each of the still image sample values in the second format comprising a sign bit, an exponent having the exponent bias value removed, and a mantissa having a second length.

24. The method of claim 23, wherein the receiving further comprises receiving a bitstream syntax element indicating a bit depth and format, and wherein the existence of the bitstream syntax element indicating the exponent bias value in the bitstream is dependent on the bitstream syntax element indicating the bit depth.

25. The method of claim 24, wherein the bitstream syntax element indicating a bit depth indicates a bit depth of 32 bits, and wherein each of the still image sample values in the second format comprise 32-bit floats.

26. The method of claim 23, wherein the still image sample values in the second format comprise high dynamic range (HDR) image data.

27. The method of claim 23, wherein the bitstream further comprises a fourth syntax element indicating whether scaling is to be performed;

and

wherein the decoding comprise, when the fourth syntax element indicates that scaling is to be performed;

(a) performing a scaling operation to reduce rounding errors during intermediate decoding operations by multiplying the still image sample values at a front end of the decoder;

(b) performing the intermediate decoding operations on the scaled still image sample values; and

(c) dividing final decoder values by an amount corresponding to the multiplying.