SYSTEMS AND METHODS FOR THE EFFICIENT COMPUTATION OF IMAGE TRANSFORMS

• US 20100303383A1
• Filed: 06/22/2009
• Published: 12/02/2010
• Est. Priority Date: 06/02/2009
• Status: Active Grant

• Current Assignee: Semiconductor Components Industries LLC (ON Semiconductor Corporation)
• Original Assignee: Aptina Imaging Corporation (ON Semiconductor Corporation)
• Inventors: Huggett, Anthony
• Granted Patent: US 8,270,763 B2
• Time in Patent Office: Days
• US Class Current: 382/282
• CPC Class Codes: G06T 3/00 Geometric image transformations in the plane of the image
• CPC Class Codes: H04N 23/80 Camera processing pipelines; Components thereof

Abstract

Image transforms that can be performed through a reverse lookup process in which output pixel addresses of an output image are reverse transformed into input pixel addresses. In some embodiments, rather than reverse transforming all output pixel addresses, a subset of output pixel addresses can be reverse transformed to calculate transformed addresses. The reverse transforms of all output pixel address may then be approximately calculated by interpolating the transformed addresses. The approximately transformed addresses can then be used as read addresses to a memory storing the input image.

Claims

1. A system comprising:

a reverse transform component operable to calculate a plurality of transformed addresses by executing a reverse transform on a subset of output pixel addresses of an output image;

a storage device operable to store the plurality of transformed addresses; and

an interpolation component operable to calculate an approximately transformed address of a particular output pixel address of the output image by interpolating at least one of the stored transformed addresses.

2. The system of claim 1, wherein the interpolation component is further operable to:

receive the particular output pixel address to be approximately transformed; and

receive from the storage device the at least one stored transformed address to be interpolated.

3. The system of claim 1, wherein the approximately transformed address is communicated as a read address to a memory device storing an input image.

4. The system of claim 1, wherein the interpolation component is further operable to calculate a particular approximately transformed address for every output pixel address of the output image.

5. The system of claim 1, wherein the storage device is further operable to store accurately transformed addresses from three rows of pixels of the output image.

6. The system of claim 5, wherein the three rows comprise:

a first row comprising an output pixel address on which the reverse transform component is currently executing a reverse transform; and

a second row comprising the at least one accurately transformed address received by the interpolation component.

7. The system of claim 1, further comprising:

a delay component operable to delay the interpolation component'"'"'s receipt of the particular output pixel address until the interpolation component can suitably calculate the approximately transformed address.

8. The system of claim 1, wherein the reverse transform component is further operable to execute the reverse transform on output pixel addresses of the subset from one row of pixels of the output image in multiple row periods of time.

9. The system of claim 1, wherein the reverse transform component is further operable to execute the reverse transform on output pixel addresses of the subset from one row of pixels of the output image in one row period of time.

10. The system of claim 1, wherein the reverse transform component comprises:

dedicated circuitry operable to perform the reverse transform.

11. The system of claim 1, wherein the reverse transform component comprises:

non-dedicated circuitry operable to execute a software program performing the reverse transform.

12. The system of claim 1, wherein the reverse transform component comprises:

an offline calculation system.

13. A method for performing an image transform comprising:

determining a subset of output pixel addresses from a plurality of output pixel addresses of an output image;

calculating a reverse transformed address (“

transformed address”

) of at least one output pixel address of the subset by performing a reverse transform, wherein the reverse transform is associated with a particular image transform; and

approximating a reverse transformed address (“

approximated address”

) of at least one of the plurality of output pixel address by interpolating the transformed address.

14. The method of claim 13, further comprising:

determining input pixel data of an input image that is associated with the approximated address; and

associating the determined input pixel data with the output pixel address from which the approximated address was approximated.

15. The method of claim 14, wherein determining input pixel data comprises:

reading input pixel data of the input image from at least two input pixel addresses neighboring the approximated address; and

interpolating the read input pixel data.

16. The method of claim 13, wherein determining a subset of output pixel addresses comprises:

defining an output pixel grid comprising a plurality of grid spaces, wherein the vertices of the grid spaces are each located at one of the plurality of output pixel addresses of the output image; and

defining the subset as the output pixel addresses located at the vertices of each grid space.

17. The method of claim 13, wherein the grid spaces comprise a particular size, and wherein the particular size creates an appropriate ratio of the quality of the image transform to the complexity of the image transform.

18. The method of claim 13, wherein:

the calculating comprises;

calculating reverse transformed addresses of output pixel addresses of the subset located on a first row and on a second row of pixels of the output image; and

the approximating comprises;

approximating the reverse transformed addresses of output pixel addresses located between the first row and the second row; and

approximating the reverse transformed addresses of output pixel addresses located on the second row.

19. The method of claim 18, wherein the calculating further comprises:

calculating reverse transformed addresses of output pixel addresses of the subset located on a third row of pixels of the output image in parallel with the approximating.

20. The method of claim 19, wherein the reverse transformed addresses calculated for the output pixel addresses of the subset located on the first row are deleted from a memory after approximating the reverse transformed addresses of output pixel addresses located between the first row and the second row.

21. A method of performing an image transform comprising:

determining a subset of output pixel addresses from a plurality of output pixel addresses of an output image;

performing a reverse transform on each output pixel address of the subset to calculate a transformed address of each output pixel address of the subset, wherein the reverse transform is associated with a particular image transform; and

for each one of the output pixel addresses, calculating an approximately transformed address for each one of the output pixel addresses by interpolating an accurately transformed address.