Data resampler for data processing system for logically adjacent data samples

• US 5,977,994 A
• Filed: 02/19/1998
• Issued: 11/02/1999
• Est. Priority Date: 10/17/1997
• Status: Expired due to Term

• Current Assignee: Microscan Systems Inc (Omron Corporation)
• Original Assignee: Acuity Brands, Inc.
• Inventors: Wilt, Michael J., Greenberg, Michael P.
• Primary Examiner(s): Chauhan, U.
• Application Number:
  US09/025,938
• Time in Patent Office: 621 Days
• Field of Search: 345/418, 345/428, 345/429, 345/437-439, 345/441-443, 345/501-504, 345/520, 345/521, 345/523, 345/507, 345/509, 345/513, 345/508
• US Class Current: 345/558
• CPC Class Codes: G06F 13/4027 using bus bridges G06F13/4022 takes precedence
• CPC Class Codes: G06T 1/60 Memory management
• CPC Class Codes: G06T 3/4007 based on interpolation, e.g. bilinear interpolation image demosaicing G06T3/4015; edge-driven or edge-based scaling G06T3/403
• CPC Class Codes: G06T 5/20 using local operators

Abstract

A data resampler for a data processing system for logically adjacent data samples is provided. The data resampler includes a memory subsystem for storing samples to be rendered, a digital differential analyzer (DDA) for generating an interpolation corner address for a sample to be rendered and which also generates a set of interpolation fractions. The resampler also includes a fetch unit, which receives the generated interpolation corner address and generates four source addresses of samples to be fetched from the memory subsystem. A number of memory units are included in the resampler. The first memory unit is a first in, first out FIFO memory, for holding the generated interpolation fractions and for permitting the DDA and fetch unit to continue to operate during memory read latency periods. The second memory unit is also a FIFO memory and is used to hold pixel data. The resampler further includes an interpolation unit, which receives pixel data from the second FIFO memory unit and interpolation fractions from the first FIFO memory unit. The interpolation unit then computes rendered result pixels, assembles the result pixels into memory words and outputs the words to a destination memory address, which is supplied by an address generator in a destination memory subsystem via a third FIFO memory unit.

Claims

1. A data resampler for a data processing system processing logically adjacent data samples, said data sampler comprising:

a memory subsystem, for storing at least a first plurality of data samples to be rendered;

a digital differential analyzer (DDA), responsive to DDA parameter sets, for generating an interpolation corner address for said plurality of data samples to be rendered, and for also generating a set of interpolation fractions;

a fetch unit, for receiving the generated interpolation corner address and for generating four source addresses of data samples to be fetched from said memory subsystem;

a first FIFO memory, for holding said interpolation fractions and for permitting said DDA and fetch unit to continue to operate during memory read latency periods;

a second FIFO memory, for holding pixel data;

a first address generator;

an address cache and a data cache, wherein said data cache comprises at least two cache lines each holding two memory words and wherein said address cache comprises at least two cache lines each comprised of two address registers, each address register for holding an address of a word in said memory subsystem; and

an interpolation unit, for receiving pixel data from said second FIFO memory and interpolation fractions from said first FIFO memory, for computing rendered result pixels, for assembling said result pixels into memory words, and for outputting said memory words to a destination memory address supplied by said address generator in a destination memory subsystem via a third FIFO.

2. The data resampler as claimed in claim 1 further including:

a data buffer, for holding DDA parameter sets in DDA parameter registers, said DDA parameter sets fetched from said memory subsystem whenever said data buffer is empty; and

a second address generator, for supplying memory subsystem addresses to said memory subsystem for DDA parameter sets to be retrieved from said memory subsystem.

3. The data resampler as claimed in claim 1, further including a fourth FIFO to extend the capacity of said data buffer, and wherein said DDA parameter registers are double-buffered and comprise two sets of registers, a first set of which supplies working values to said DDA and a second set of which is available to be loaded from said fourth FIFO under control of a sequencer in order to eliminate parameter loading delay.

4. The data resampler as claimed in claim 1, wherein the width of said first FIFO is increased to carry a data descriptor along with each set of interpolation fractions.

5. The data resampler as claimed in claim 1 further comprising an adder for generating a lower-left pixel address from an upper-left pixel address by adding a stride to said upper-left pixel address.

6. The data resampler as claimed in claim 1 further comprising four working address registers organized as two address cache lines each having left and right registers for holding addresses of memory words, wherein in each line the left register holds the address of the memory word containing the left pixel, and the right register holds the address of the next word.

7. The data resampler as claimed in claim 6 further comprising an incrementer for generating correct memory addresses whenever a cache line is updated.

8. The data resampler as claimed in claim 7 further comprising at least one comparator for comparing the memory addresses of the upper-left and lower-left pixels with the contents of all four working address registers.

9. The data resampler as claimed in claim 8, further comprising a data descriptor encoder and a data switch and sequencer, said descriptor encoder for generating a data cache update descriptor and for instructing said data switch and sequencer to update the address cache working registers and to request words to be fetched from memory via said fetch unit.

10. The data resampler as claimed in claim 1, wherein said data cache comprises four working registers each holding a memory word, said registers arranged in left and right pairs, each pair comprising a data cache line holding twice the number of pixels per memory word.

11. The data resampler as claimed in claim 10, further including a second data switch and sequencer, which updates working registers based on the data descriptor from the output of said first FIFO and a pixel output multiplexer.

12. A method of resampling data for a data processing system for logically adjacent data samples, said method comprising:

storing at least data samples to be rendered in a memory subsystem;

generating, in response to digital differential analyzer (DDA) parameter sets stored in DDA parameter set registers, an interpolation corner address for a sample to be rendered and a set of interpolation fractions using a digital differential analyzer (DDA);

receiving the generated interpolation corner address and responsive to said received interpolation corner address, generating four source addresses of samples to be fetched from said memory subsystem using an address cache;

comparing said four source addresses with all addresses resident in four working address registers of said address cache;

determining if any new address needs to be fetched from said memory subsystem and fetching any required address from said memory subsystem;

holding said interpolation fractions in a FIFO memory to permit said DDA to continue to operate during memory read latency periods;

holding pixel data in a second FIFO; and

receiving pixel data from said second FIFO and interpolation fractions from said first FIFO and computing rendered result pixels, assembling said result pixels into memory words, outputting said words to a destination memory address supplied by an address generator in a destination memory subsystem via a third FIFO, and reloading DDA parameter set registers after each DDA rendering cycle.

13. The method of resampling data as claimed in claim 12, further comprising pre-fetching blocks of parameters into a data buffer.

14. The method of resampling data as claimed in claim 13 further comprising double-buffering said parameter registers into two sets of registers, a first register set of said two sets of registers of which supplies working values to said DDA, and a second register set of which is available to be loaded from said fourth FIFO under control of a sequencer in order to eliminate parameter loading delay.

15. The method of resampling data as claimed in claim 14 further comprising providing a data cache having two cache lines and holding two memory words in each of said cache lines.

16. The method of resampling data as claimed in claim 12 further comprising encoding and outputting a data descriptor, which specifies how the data cache is to update it'"'"'s registers from said second FIFO.

17. The method of resampling data as claimed in claim 16, further comprising fetching two memory words whenever a cache line is re-filled.

18. The method of resampling data as claimed in claim 17, wherein said data descriptor also instructs said data cache to load the upper cache line from the lower cache line.

19. The method of resampling data as claimed in claim 17, wherein said data descriptor also instructs said data cache to load the lower cache line from the upper cache line.

20. The method of resampling data as claimed in claim 12, wherein a lower-left pixel address is generated by adding the stride to the upper-left (corner) pixel address using an adder.

21. The method of resampling data as claimed in claim 12, further comprising organizing the four working address registers as two address cache lines each containing left and right address registers holding the address of the memory word containing the left pixel in the left register and holding the address of the next word in the right address register.

22. The method of resampling data as claimed in claim 21, further comprising generating correct memory addresses whenever a cache line is updated using an incrementer.

23. The method of resampling data as claimed in claim 22, further comprising comparing the memory address of the upper-left and lower-left pixels with the contents of all four working address registers.

24. The method of resampling data as claimed in claim 23, further comprising generating a data cache update descriptor and instructing a data switch and sequencer to update the address cache working registers and requesting words to be fetched from memory.