Intra-frame quantizer selection for video compression

• US 6,256,423 B1
• Filed: 12/15/1998
• Issued: 07/03/2001
• Est. Priority Date: 09/18/1998
• Status: Expired due to Term

• Current Assignee: LG Electronics, Inc. (LG Corporation), SRI International, Inc.
• Original Assignee: LG Electronics, Inc. (LG Corporation), Sarnoff Corporation (SRI International, Inc.)
• Inventors: Krishnamurthy, Ravi, Sethuraman, Sriram
• Primary Examiner(s): Tran, Phucc
• Assistant Examiner(s): Alavi, Amir
• Application Number:
  US09/212,025
• Time in Patent Office: 931 Days
• Field of Search: 348/401, 348/402, 348/403, 348/404, 382/248, 382/249, 382/250, 382/251, 382/252
• US Class Current: 382/251
• CPC Class Codes: H04N 19/126 Details of normalisation or weighting functions, e.g. normalisation matrices or variable uniform quantisers
• CPC Class Codes: H04N 19/149 by estimating the code amount by means of a model, e.g. mathematical model or statistical model
• CPC Class Codes: H04N 19/15 by monitoring actual compressed data size at the memory before deciding storage at the transmission buffer
• CPC Class Codes: H04N 19/152 by measuring the fullness of the transmission buffer
• CPC Class Codes: H04N 19/154 Measured or subjectively estimated visual quality after decoding, e.g. measurement of distortion use of rate-distortion criteria H04N19/147
• CPC Class Codes: H04N 19/162 User input
• CPC Class Codes: H04N 19/17 the unit being an image region, e.g. an object
• CPC Class Codes: H04N 19/192 the adaptation method, adaptation tool or adaptation type being iterative or recursive
• CPC Class Codes: H04N 19/196 being specially adapted for the computation of encoding parameters, e.g. by averaging previously computed encoding parameters processing of motion vectors H04N19/513
• CPC Class Codes: H04N 19/198 including smoothing of a sequence of encoding parameters, e.g. by averaging, by choice of the maximum, minimum or median value
• CPC Class Codes: H04N 19/30 using hierarchical techniques, e.g. scalability H04N19/63 takes precedence
• CPC Class Codes: H04N 19/60 using transform coding

Abstract

An image is divided into one or more (e.g., foreground) regions of interest with transition regions defined between each region of interest and the relatively least-important (e.g., background) region. Each region is encoded using a single selected quantization level, where quantizer values can differ between different regions. In general, in order to optimize video quality while still meeting target bit allocations, the quantizer assigned to a region of interest is preferably lower than the quantizer assigned to the corresponding transition region, which is itself preferably lower than the quantizer assigned to the background region. The present invention can be implemented iteratively to adjust the quantizer values as needed to meet the frame'"'"'s specified bit target. The present invention can also be implemented using a non-iterative scheme that can be more easily implemented in real time. The present invention enables a video compression algorithm to meet a frame-level bit target, while ensuring spatial and temporal smoothness in frame quality, thus resulting in improved visual perception during playback.

Claims

1. A method for processing digital image data of an already digitized image in accordance with a compression algorithm that imposes a maximum difference between quantization levels for consecutive macroblocks in the digitized image, comprising the steps of:

(a) identifying one or more macroblocks corresponding to a region of interest in the digitized image;

(b) identifying one or more macroblocks corresponding to a transition region in the digitized image located between the region of interest and a least-important region in the digitized image;

(c) selecting a first quantization level for each macroblock in the region of interest;

(d) selecting a second quantization level for each macroblock in the transition region, wherein the second quantization level is greater than the first quantization level;

(e) selecting a third quantization level for each macroblock in the least-important region, wherein;

the third quantization level is greater than the second quantization level; and

the difference between the third and first quantization levels is up to twice as large as the maximum difference imposed by the compression algorithm; and

(f) encoding the digitized image based on the selected first, second, and third quantization levels in accordance with the compression algorithm.

2. The invention of claim 1, further comprising the steps of:

(g) comparing the number of bits used to encode the image in step (f) to a bit target for the image;

(h) adjusting one or more of the first, second, and third quantization levels in accordance with the comparison of step (g); and

(i) re-encoding the image based on the adjusted quantization levels.

3. The invention of claim 2, wherein steps (g)-(i) are repeated until the number of bits used to encode the image is sufficiently close to the bit target.

4. The invention of claim 2, wherein:

if the number of bits in step (g) is sufficiently low, then step (h) comprises the step of decreasing the first quantization level and, if appropriate, decreasing the second quantization level, and then, if appropriate, decreasing the third quantization level; and

if the number of bits in step (g) is sufficiently high, then step (h) comprises the step of increasing the third quantization level and, if appropriate, increasing the second quantization level, and then, if appropriate, increasing the first quantization level.

5. The invention of claim 1, wherein the image has two or more regions of interest and each region of interest is assigned its own quantization level, which may differ between regions of interest.

6. The invention of claim 1, wherein the region of interest corresponds to a talking head and the least-important region corresponds to a relatively stationary background.

7. The invention of claim 1, wherein at least one of the first, second, and third quantization levels is selected based on modeling of rate-distortion curves at different quantization levels.

8. The invention of claim 7, wherein a number of bits used to encode a macroblock is modeled as follows:

9. The invention of claim 8, wherein the model is initialized between an I frame and the following P frame by encoding each macroblock of an intervening frame using two or more quantization levels.

10. The invention of claim 8, wherein the model is updated as the image data are coded.

11. A computer-readable medium having stored thereon a plurality of instructions, the plurality of instructions including instructions which, when executed by a processor, cause the processor to implement a method for processing digital image data of an already digitized image in accordance with a compression algorithm that imposes a maximum difference between quantization levels for consecutive macroblocks in the digitized image, the method comprising the steps of:

(a) identifying one or more macroblocks corresponding to a region of interest in the digitized image;

(b) identifying one or more macroblocks corresponding to a transition region in the digitized image located between the region of interest and a least-important region in the digitized image;

(c) selecting a first quantization level for each macroblock in the region of interest;

(d) selecting a second quantization level for each macroblock in the transition region, wherein the second quantization level is greater than the first quantization level;

(e) selecting a third quantization level for each macroblock in the least-important region, wherein;

the third quantization level is greater than the second quantization level; and

the difference between the third and first quantization levels is up to twice as large as the maximum difference imposed by the compression algorithm; and

(f) encoding the digitized image based on the selected first, second, and third quantization levels in accordance with the compression algorithm.

12. The invention of claim 11, further comprising the steps of:

(g) comparing the number of bits used to encode the image in step (f) to a bit target for the image;

(h) adjusting one or more of the first, second, and third quantization levels in accordance with the comparison of step (g); and

(i) re-encoding the image based on the adjusted quantization levels.

13. The invention of claim 12, wherein steps (g)-(i) are repeated until the number of bits used to encode the image is sufficiently close to the bit target.

14. The invention of claim 12, wherein:

if the number of bits in step (g) is sufficiently low, then step (h) comprises the step of decreasing the first quantization level and, if appropriate, decreasing the second quantization level, and then, if appropriate, decreasing the third quantization level; and

if the number of bits in step (g) is sufficiently high, then step (h) comprises the step of increasing the third quantization level and, if appropriate, increasing the second quantization level, and then, if appropriate, increasing the first quantization level.

15. The invention of claim 11, wherein the image has two or more regions of interest and each region interest is assigned its own quantization level, which may differ between regions of interest.

16. The invention of claim 11, wherein the region of interest corresponds to a talking head and the least-important region corresponds to a relatively stationary background.

17. The invention of claim 11, wherein at least one of the first, second, and third quantization levels is selected based on modeling of rate-distortion curves at different quantization levels.

18. The invention of claim 17, wherein a number of bits used to encode a macroblock is modeled as follows:

19. The invention of claim 18, wherein the model is initialized between an I frame and the following P frame by encoding each macroblock of an intervening frame using two or more quantization levels.

20. The invention of claim 18, wherein the model is updated as the image data are coded.