Adaptive quantization of depth signal in 3D visual coding

• US 7,558,432 B2
• Filed: 04/21/2005
• Issued: 07/07/2009
• Est. Priority Date: 04/29/2004
• Status: Expired due to Fees

• Current Assignee: Mitsubishi Electric Corporation
• Original Assignee: Mitsubishi Electric Corporation
• Inventors: Zaharia, Ramona, Bober, Miroslaw, Cieplinski, Leszek
• Primary Examiner(s): Strege; John B
• Application Number:
  US11/110,764
• Publication Number:
  US 20050244071A1
• Time in Patent Office: 1,538 Days
• Field of Search: 382/251, 382/154, 382/239, 382/168
• US Class Current: 382/251
• CPC Class Codes: G06T 9/007 Transform coding, e.g. discrete cosine transform
• CPC Class Codes: G06T 9/008 Vector quantisation
• CPC Class Codes: H04N 13/128 Adjusting depth or disparity
• CPC Class Codes: H04N 13/221 using the relative movement between cameras and objects
• CPC Class Codes: H04N 19/126 Details of normalisation or weighting functions, e.g. normalisation matrices or variable uniform quantisers
• CPC Class Codes: H04N 19/597 specially adapted for multi-view video sequence encoding

Abstract

A method of representing an image or sequence of images using a depth map comprises transforming an n-bit depth map representation into an m-bit depth map representation, where m<n, using a non-linear transformation.

Claims

1. A method of representing an image or sequence of images using a depth map comprising:

using one or more processors to perform the steps of;

requantizing an n-bit depth map representation into an m-bit depth map representation, where m<

n, using a non-linear transformation, comprising transforming a depth map in the form of an n-bin histogram into a second m-bin histogram, the widths of at least two bins in the m-bin histogram being unequal,

wherein the transforming comprises;

predetermining the m bins of the second histogram; and

combining bins of the first n-bin histogram by mapping pixels from one or more bins of the first histogram into a bin of the second histogram.

2. A method as claimed in claim 1 wherein more bins in the second histogram correspond to bins of the first histogram having higher histogram count.

3. A method as claimed in claim 1 wherein approximately the same number of pixels is assigned to each bin of the m-bit histogram.

4. A method as claimed in claim 1 wherein the non-linear transformation is a piecewise linear transformation, each piecewise linear transformation mapping one or more bins of the first histogram into a bin of the second histogram.

5. A method as claimed in claim 1 comprising mapping until the number of pixels in the bin of the second histogram exceeds a predetermined threshold.

6. A method as claimed in claim 5 wherein the threshold TH=Width×

Height/2^m where width and height represent the dimensions of the image.

7. A method as claimed in claim 1 comprising grouping pixels into bins using an error measurement.

8. A method as claimed in claim 7 wherein the error value is the sum of squared differences between the pixels in the bin and a representative value of the bin.

9. A method as claimed in claim 1, wherein the transformation is recursive and involves intermediate transformations to one or more p-bin histograms, where p is variable.

10. A method as claimed in claim 1 using an algorithm based on a Lloyd-Max quantizer.

11. A method as claimed in claim 1 wherein m=8.

12. A method as claimed in claim 1 wherein for different areas of an image, different transformations or different quantization steps are used, based on a predetermined criterion.

13. A method as claimed in claim 1 for representing a sequence of images wherein said representation is derived for each of a plurality of images in the sequence.

14. A method as claimed in claim 1 for representing a sequence of images wherein said representation is derived for a plurality of images.

15. A method as claimed in claim 14 comprising combining depth map histograms for the plurality of images and deriving the m-bit depth map representation from the combined histogram.

16. A method as claimed in claim 1 for representing a sequence of images comprising comparing the depth map for a first image with the depth map for a second image and determining the representation for the first image based on the comparison.

17. Control device programmed to perform a method as claimed in claim 1.

18. A computer program embodied on a computer-readable medium, the computer program for executing a method as claimed in claim 1.

19. A method as claimed in claim 8, wherein the representative value of the bin is a representative weighted value of the bin.

21. Apparatus as claimed in claim 20 comprising one or more of storage means storing images and/or descriptors of images, display means for displaying images, input means for inputting images, coding and/or decoding means, transmitting and/or receiving means for transmitting and/or receiving images, processing means, storage means storing instructions for executing a method of claim 1.

20. Apparatus for representing an image or sequence of images using a depth map comprising:

a requantizing unit for requantizing an n-bit depth map representation into an m-bit depth map representation, where m<

n, using a non-linear transformation, comprising a transforming unit for transforming a depth map in the form of an n-bin histogram into a second m-bin histogram, the widths of at least two bins in the m-bin histogram being unequal,

wherein the transforming unit is adapted to;

predetermine the m bins of the second histogram; and

combine bins of the first n-bin histogram by mapping pixels from one or more bins of the first histogram into a bin of the second histogram.