Memory efficient on-chip buffering for projective transformation

• US 10,152,767 B1
• Filed: 12/15/2016
• Issued: 12/11/2018
• Est. Priority Date: 12/15/2016
• Status: Active Grant

• Current Assignee: The MathWorks, Inc.
• Original Assignee: The MathWorks, Inc.
• Inventors: Hanson, Michael Neil, Rice, Garrey William, Ogilvie, Brian K., Jachimczyk, Witold R.
• Primary Examiner(s): Nguyen, Hau H
• Application Number:
  US15/380,500
• Time in Patent Office: 726 Days
• CPC Class Codes: G06F 3/0673 Single storage device
• CPC Class Codes: G06T 2200/28 involving image processing hardware
• CPC Class Codes: G06T 3/00 Geometric image transformations in the plane of the image

Abstract

A device may receive information associated with an input image that includes a first quantity of rows. The device may store pixel values, for a second quantity of rows, based on the information associated with the input image. The second quantity of rows may be different than the first quantity of rows. The device may generate an output image, that includes a projective transformation of the input image, based on the stored pixel values. The device may provide information associated with the output image.

Claims

1. A method, comprising:

receiving, by one or more devices, information associated with an input image that includes input pixels arranged in input rows and input columns within a set of input coordinates,

the input image to be transformed into an output image that includes output pixels arranged in output rows and output columns within a set of output coordinates;

determining, by the one or more devices or one or more different devices, a mapping between a first input pixel in a first input row, of the input rows, and a corresponding output pixel in a first output row of the output rows;

determining, by the one or more devices or the one or more different devices, a quantity of rows based on the mapping,

the quantity of rows being smaller than a total number of the input rows;

selectively storing pixel values that are in or between a starting input row and an end input row,

a total number of rows from the starting input row to the end input row being the quantity of rows;

determining a row replication value that indicates a quantity of times that the first output row or the first input row is to be replicated in the output image; and

generating the output image using at least some of the stored pixel values,

the generating the output image comprising determining that the row replication value is satisfied, and

the output image being a transformed image of the input image.

2. The method of claim 1, where selectively storing the pixel values comprises:

storing the pixel values using on-chip random access memory.

3. The method of claim 1, where selectively storing the pixel values comprises:

storing a first quantity of pixel values associated with the first input row; and

storing a second quantity of pixel values associated with a second row of the input rows, the first quantity being different than the second quantity.

4. The method of claim 1, further comprising:

determining a stored row based on an output row;

determining a row start address associated with the stored row;

determining a set of pixel values, associated with a buffer, based on the row start address and a gradient value; and

performing an interpolation technique in association with the set of pixel values,

where generating the output image comprises;

generating the output image based on performing the interpolation technique in association with the set of pixel values.

5. The method of claim 1,

where the row replication value is associated with the starting input row;

and

where the method further comprises;

incrementing an accumulator value based on the row replication value being satisfied,

the accumulator value being incremented by a first quantity of pixel values,

the first quantity of pixel values being associated with the starting input row, and

where generating the output image comprises;

generating the output image based on the accumulator value.

6. The method of claim 1, where the mapping is determined based on information associated with an image acquisition device used to generate the input image.

7. The method of claim 1,

where the output image is a bird'"'"'s eye view image, and

where the input image is not a bird'"'"'s eye view image.

8. A non-transitory computer-readable medium storing instructions, the instructions comprising:

one or more instructions that, when executed by one or more processors, cause the one or more processors to;

identify information associated with an input image that includes input pixels arranged in input rows and input columns within a set of input coordinates,

the input image to be transformed into an output image that includes output pixels arranged in output rows and output columns within a set of output coordinates;

determine a mapping between a first input pixel in a first input row, of the input rows, and a corresponding output pixel in a first output row of the output rows;

determine a quantity of rows based on the mapping;

selectively store pixel values that are in or between a starting input row and an end input row,

a total number of rows from the starting input row to the end input row being the quantity of rows;

determine a row replication value that indicates a quantity of times that the first output row or the first input row is to be replicated in the output image; and

generate the output image based on the stored pixel values,

where the one or more instructions that cause the one or more processors to generate the output image comprise one or more instructions that cause the one or more processors to determine that the row replication value is satisfied, and

the output image being a transformed image of the input image.

9. The non-transitory computer-readable medium of claim 8, where the one or more instructions, when executed by the one or more processors, cause the one or more processors to:

determine the first input pixel that maps to the corresponding output pixel; and

determine a starting input row value based on the first input pixel.

10. The non-transitory computer-readable medium of claim 8, where the row replication value is determined based on the mapping.

11. The non-transitory computer-readable medium of claim 8, where the one or more instructions, when executed by the one or more processors, cause the one or more processors to:

determine a first start column value for the first input row; and

determine a second start column value for a second input row of the input rows,

the second start column value being different than the first start column value.

12. The non-transitory computer-readable medium of claim 8,

where the one or more instructions, when executed by the one or more processors, cause the one or more processors to:

receive information associated with an image acquisition device,

the image acquisition device to generate the input image, and

where the mapping is determined based on the information associated with the image acquisition device.

13. The non-transitory computer-readable medium of claim 8,

where the output image is a bird'"'"'s eye view image, and

where the input image is not a bird'"'"'s eye view image.

14. A device, comprising:

a memory; and

one or more processors to;

receive first information associated with an input image that includes input pixels arranged in input rows and input columns within a set of input coordinates,

the input image to be transformed into an output image that includes output pixels arranged in output rows and output columns within a set of output coordinates;

determine a mapping between a first input pixel in a first input row, of the input rows, and a corresponding output pixel in a first output row of the output rows;

determine a quantity of rows based on the mapping,

the quantity of rows being smaller than a total number of the input rows;

selectively store pixel values that are in or between a starting input row and an end input row,

a total number of rows from the starting input row to the end input row being the quantity of rows;

determine a row replication value that indicates a quantity of times that the first output row or the first input row is to be replicated in the output image; and

generate the output image using at least some of the stored pixel values,

when generating the output image, the one or more processors are to

determine that the row replication value is satisfied, and

the output image being a transformed image of the input image.

15. The device of claim 14, where the one or more processors are further to:

store a different number of pixel values for each row of the input image.

16. The device of claim 14,

where the output image is a bird'"'"'s eye view image, and

where the input image is not a bird'"'"'s eye view image.

17. The device of claim 14,

where the one or more processors are further to:

determine a start row address, associated with a second stored row, based on a quantity of pixel values in the set of input coordinates; and

where the output image is generated based on the start row address.

18. The device of claim 14,

where the one or more processors are further to:

increment an accumulator value based on a third quantity of pixel values associated with the first output row; and

where the output image is generated based on the accumulator value.

19. The device of claim 14, where the one or more processors are further to:

receive second information that identifies a first gradient value associated with a first stored row; and

receive third information that identifies a second gradient value associated with a second stored row,

the first gradient value and the second gradient value being different; and

where the one or more processors, when generating the output image, are to;

generate the output image based on the first gradient value and the second gradient value.

20. The device of claim 14, where the mapping is determined based on information associated with an image acquisition device used to generate the input image.