Digital image warping system

Abstract

A system is provided for performing real-time warping in one or two dimensions (eg. dynamic zoom and shrink in both horizontal and vertical directions), and temporal warping of image frames (eg. fast motion and slow motion). The system accomplishes warping with proper Nyquist bandlimiting. Three different processes are performed in sequence to create the warped image according to the present invention. The first process maps a desired target pixel or frame spacing to an appropriate FIR filter wherein the size of the FIR filter is chosen based on the desired target pixel or frame spacing. Once the filter is chosen, a second process creates an appropriate number of upsampled pixels, lines or frames based on the input source pixels, lines or frames. The last process multiplies the correct filter coefficients from the first process with the interpolated pixels, lines or frames from the second process, resulting in correctly Nyquist bandlimited target pixels, lines or frames.

Description

FIELD OF THE INVENTION

The present invention relates in general to digital signal processing, and more particularly to an image warping system for generating an output stream of warped pixels and/or lines and/or frames from an input stream of pixels and/or lines and/or frames, respectively.

BACKGROUND OF THE INVENTION

Digital image warping is a process of dynamically resampling a regularly spaced source image to produce a target image with different spacing than the source image. Many modern movies use digital warping to create mind boggling effects such as perspectives, odd shaped image boundaries, and texture mapping, to name but a few. Prior art real time image warping systems are extremely complicated and expensive due to the massive number of computations which must be performed in real time. Prior art systems do not dynamically bandlimit the spatial frequency of the warped image to the Nyquist limit, causing unwanted aliasing distortion. Examples of well known prior art digital image warping systems are described in the following publications and patent:

I. Wolberg, George, "Separable Image Warping: Implications and Techniques", Ph.D. Thesis, Dept. of Computer Science, Columbia University, N.Y., 1990, and "Digital Image Warping", IEEE Computer Society Press, Los Alamitos, Calif., 1990; II. U.S. Pat. No. 5,355,328 (Arbeiter, et al).

SUMMARY OF THE INVENTION

According to the present invention, an image warping system is provided for performing real-time dynamic zoom and shrink in both horizontal and vertical directions with proper Nyquist bandlimiting, and requiring no external line stores or other memory. This same invention can also temporally warp images (eg. dynamic changes between fast motion and slow motion) by utilizing external frame stores in place of registers. Accordingly, the system of the present invention constitutes a significant advance over the known prior art. Three different processes are performed in sequence to create the warped image according to the present invention. The first process maps a desired target pixel (or line or frame) spacing to an appropriate FIR filter wherein the size of the FIR filter is chosen based on the desired target pixel (or line or frame) spacing. Once the filter is chosen, a second process creates an appropriate number of upsampled pixels (or lines or frames) based on the input source pixels (or lines or frames). The last process multiplies the correct filter coefficients from the first process with the interpolated pixels (or lines or frames) from the second process, resulting in correctly Nyquist bandlimited target pixels (or lines or frames).

Thus, in accordance with one aspect of the invention, there is provided an apparatus and method for digitally warping an input stream of one of either pixels and/or lines and/or frames to produce an output stream of warped pixels and/or lines and/or frames in accordance with a user defined target increment defining the desired spacing between output pixels and/or lines and/or frames. In the case of digitally warped pixels and/or lines, the resulting image is considered to be spatially warped. In the case of digitally warped frames, the resulting set of frames is considered to be temporally warped.

In a first embodiment of the invention, the system comprises a filter look-up table for mapping the target increment into a specific filter size, and a pair of registers to pipeline three successive filter sizes output from the filter look-up table for input to a pair of comparators. A first one of the comparators determines the largest filter size from the three values output by the filter look-up table and both registers, while the second comparator determines the largest filter size output from the filter look-up table and the first register. The output of the first comparator is received by an upsample increment generator which maps the filter size to a filter factor and then right-shifts the target increment by this amount to generate an upsample increment. The upsample increment is received by an interpolator which in response generates an appropriate number of evenly spaced upsampled pixel values. The output from the second comparator is stored in a pipeline comprising two further registers connected respectively to left and right half kernel filter generators. The left and right half kernel filter generators receive and multiply the upsampled pixels from the interpolator by respective half kernel coefficients to create the warped output pixels.

In a second embodiment of the invention, the system comprises a filter look-up table for mapping the target increment into a specific filter size, and a single register for holding the filter size from the filter look-up table for input into left and right half kernel filter generators. The output of the filter look-up table is received by an upsample increment generator which maps the filter size to a filter factor and then right-shifts the target increment by this amount to generate an upsample increment. The upsample increment is received by an interpolator which in response generates an appropriate number of evenly spaced upsampled pixel values. The left and right half filter kernel generators receive and multiply the upsampled pixels from the interpolator by respective half kernel coefficients to create the warped output pixels.

Each embodiment can be extended to digitally warp lines and/or frames by replacing the registers with line stores and frame stores, respectively. These larger memory storage units perform the identical function as the registers in the embodiments described in detail below.

The warping technique of the present invention may be implemented utilizing the image filtering and image extension apparatus and methods described in co-pending U.S. patent application Ser. Nos. 08/126,388; 08/125,530; 08/172,065; 08/124,201, all assigned to Genesis Microchip Inc., the contents of which are incorporated herein by reference.

Furthermore, the digital image warping system according to the present invention builds on technology described in U.S. Pat. No. 5,355,328 (Arbeiter et al), referred to above, which relates to a method of resizing source images to an arbitrary but fixed target image size. According to this prior art system, each target pixel and/or line and/or frame is resized by the same amount relative to the source pixel and/or line and/or frame. According to the present invention, each target pixel and/or line and/or frame may be individually sized arbitrarily relative to the source pixel and/or line and/or frame. More specifically, according to the present invention, source pixels and/or lines and/or frames are dynamically upsampled based on changing target pixel and/or line and/or frame resize factors. The source pixels and/or lines and/or frames are always upsampled to the nearest first octave and then decimated by fixed sized FIR filters to create Nyquist bandlimited target pixels and/or lines and/or frames. The filter sizes are dynamically determined for each target pixel and/or line and/or frame based on spatial positioning or temporal positioning as dictated by the user. Major silicon cost reduction is achieved with the present invention over prior art video warping systems through the use of fixed size FIR filters.

BRIEF DESCRIPTION OF THE DRAWINGS

A detailed description of the preferred embodiment is provided herein below with reference to the following drawings, in which:

FIG. 1 shows a plurality of target pixels derived from a plurality of source pixels, wherein the target pixels are spaced according to the user defined target increment;

FIG. 2 is a block diagram showing functional components of the digital image warping system according to the present invention;

FIG. 3 shows the relative spatial positioning of source and target pixels for dynamically varying user defined target increments, and the selection of appropriate FIR filters for generating the target pixels using the system of FIG. 2, according to a representative example;

FIG. 4 is a block diagram showing functional components of the digital image warping system according to the preferred embodiment of the present invention; and

FIG. 5 shows the relative spatial positioning of the source and target pixels for dynamically varying user defined target increments, and the selection of appropriate FIR filters for generating the target pixels using the preferred embodiment of FIG. 4.

DETAILED DESCRIPTION OF THE ALTERNATIVE AND PREFERRED EMBODIMENTS

In the description below, the preferred and alternative embodiments apply equally to the warping of pixels and/or lines. The same inventive principles can be used to temporally warp frames by substituting frames stores for each register in FIGS. 2 and 4. Furthermore, the invention is not limited to video image warping. For example, in the case of one-dimensional warping, the principles of the present invention may be applied to audio rather than video samples (eg. to correct jitter on a digital audio tape drive).

As discussed above, digital warping of an image is specified by target pixel spacing with respect to source pixel spacing. The horizontal space between each target pixel and the vertical space between each target line must be specified by the user. Likewise, the temporal spacing (ie. time) between successive frames of an image must be user specified. This detailed information provides a very fine level of control that can be used to create very complex warps. The generation of user defined target spacing is known in the art, and does not form part of the present invention. The system of the present invention receives as inputs a stream of source pixels (and/or lines and/or frames) and the desired target increment (TarInc) which defines the user-specified spacing.

FIG. 1 illustrates the generation of target pixels from source pixels based on a user defined target increment (TarInc). For simplicity, the remainder of this disclosure will refer to the generation of target pixels from source pixels, it being understood that exactly the same methodology and circuitry applies to vertical processing by substituting video lines for pixels and temporal processing by substituting video frames for pixels.

As can be seen from FIG. 1, the first and last target pixels of a warped image are spatially aligned with the first and last source pixels. However, the spacing of the intermediate target pixels is defined by the target increment (TarInc). Specifically, in the example shown in FIG. 1, the first (i.e. left-most) intermediate target pixel is spaced from the first target pixel by an amount equal to 1.5 times the spacing between the source pixels, the second intermediate target pixel is spaced from the first intermediate target pixel by an amount equal to 0.75 times the spacing between the source pixels, while the last target pixel is separated from the second intermediate target pixel by an amount equal to 0.75 times the spacing between the source pixels, for a total spacing of 3.0 (i.e. the spacing between the first and last source pixels).

Turning now to FIG. 2, a functional block diagram is shown according to a first embodiment of the invention for generating a stream of warped output pixels from a stream of source pixels in accordance with dynamically varying target increments.

Firstly, the target increment (TarInc) is mapped via filter look-up table 21 to a specific FIR filter size as represented by a filter factor value based on the mapping shown in Table A, below. The FIR coefficients can be stored in a read-only table (i.e. ROM) or can be supplied by the user. Programming the coefficients gives the user the flexibility to tune the FIR filters. The filter size is chosen by indexing the filter look-up table based on TarInc. It is important to note that Table A is only an example of one implementation. This table could be extended to support any size of TarInc and generate any filter size and filter factor.

TABLE A______________________________________FIR Filter Table and Filter FactorTarInc(t) 0<t<1 1<t<2 2<t<4 4<t<8 8<t<16 16<t<32______________________________________Filter 3 Tap 5 Tap 9 Tap 17 Tap 33 Tap 65 TapSizeFilter 0 1 2 3 4 5Factor______________________________________

Once TarInc is mapped, the largest filter on either side of a target pixel is chosen as the filter for that pixel. FIG. 3 illustrates this step. The fifth line of FIG. 3 (TarInc.RTM. Filter Mapping) shows the TarInc values from the fourth line (Target Increment) mapped to various FIR filters based on the mapping of Table A, above. This is represented schematically in FIG. 3 by placing the largest filter size (i.e. number of taps) of the filter sizes appearing on each side of the target pixel inside the target pixel, indicating that it is the chosen filter for that pixel. In FIG. 3, the third line (Target) shows the first target pixel being generated by a 9 Tap filter, the second target pixel being generated by a 9 Tap filter, the third target pixel being generated by a 3 Tap filter, etc.

As discussed above, TarInc is mapped to an FIR filter size according to the mapping Table A, via filter look-up table 21, the output of which (i.e. filter factor value) is stored successively in A and B registers 22 and 23, respectively, on successive clock cycles, thereby forming a pipeline. Comparator 24 determines which of two adjacent filters on each side of a target pixel is the larger, and stores a representation on this filter (i.e. the associated filter factor) in the left register 25 connected to the output of comparator 24. The left kernel MAC 26 (i.e. multiplier accumulator) reads the filter factor value stored in left register 25 and generates half of the target pixel value using the left half of the designated filter kernel. On the next clock cycle, the right kernel MAC 27 generates the rest of the target pixel value using the right half of the filter kernel and the centre tap stored in right register 28, and sums the resulting value with the output value received from left kernel MAC 26, resulting in the warped output pixel.

Before dynamic filtering can take place as discussed above in connection with comparator 24, the left and right registers 25 and 28, and left and right kernel MACs 26 and 27, the correct number of upsampled pixels must be generated. Referring again to FIG. 2, the number of upsampled pixels is determined by the largest filter size stored before and after A and B registers 22 and 23.

Comparator 29 determines the largest filter size (i.e. largest filter factor output from filter look-up table 21) and transmits that value to an upsample increment generator 30. The upsample increment generator 30 receives and right-shifts the TarInc value by a predetermined number of bits equivalent to the received filter factor to produce an upsample increment value. For the example of FIG. 3, the first target increment is 2.25, resulting in a filter factor of two which corresponds to a "divide-by-four" operation (i.e. right-shift TarInc by two bits resulting in an upsample increment of 0.5625). The upsample increment value (i.e. right-shifted TarInc value) represents the correct equal spacing of upsampled pixels for application to the required size of FIR filter (e.g. 9-Tap filter for TarInc=2.25).

Interpolator 31 reads the upsample increment value and interpolates upsampled pixels from the source pixels on a continuous basis and with the spacing defined by the upsample increment value, until an internal incrementor or counter value reaches TarInc. For the example discussed above and shown in FIG. 3, four upsampled pixels are generated with the spacing of 0.5625 relative to the source pixel spacing, for application to the 9 Tap FIR filter. The interpolator 31 may be implemented as a well known linear interpolator, a quadratic interpolator as disclosed in applicant's co-pending application Ser. No. 08/172,065, or any other suitable design of interpolator.

The upsampled pixels generated by interpolator 31 are transmitted to the left and right kernel MACs 26 and 27 which contain control logic to ensure that the correct coefficient is multiplied by the correct upsampled pixel. Thus, in FIG. 3, line 2 (Upsample) depicts the largest possible number of required upsampled pixels for the different sized FIR filters implemented to generate the target pixels. The FIR filters shown operating between lines 2 and 3 of FIG. 3 are implemented by left and right half kernel MACs 26 and 27 which multiply the correct coefficients by the correct upsampled pixel values. The internal control logic of the left and right kernel MACs 26 and 27 ensures that predetermined ones of the upsampled pixels are skipped for correct matching of coefficients and upsampled pixels. For example, whereas the 9 Tap FIR filter utilized to generate the first intermediate target pixel utilizes all of the upsampled pixels from line 2, the adjacent 3 Tap filter utilizes only the fifth, ninth and eleventh generated upsampled pixels, and skips the sixth, seventh, eighth and tenth upsampled pixels, etc.

The control logic of left and right kernel MACs 26 and 27 also generates the first target pixel on start up of the system by doubling the right half kernel values for the first target pixel, as described in our copending application Ser. No. 08/126,388. The last target pixel is generated by doubling the left kernel values. This doubling is required to ensure that all FIR coefficients maintain unity gain.

Turning now to FIG. 4, a functional block diagram is shown according to the preferred embodiment for generating a stream of warped output pixels from a stream of source pixels in accordance with dynamically varying target increments.

Firstly, the target increment (TarInc) is mapped via filter look-up table 41 to a specific FIR filter size as represented by a filter factor value based on the mapping shown in Table A, above. The FIR coefficients can be stored in a read-only table (i.e. ROM) or supplied by the user. The filter size is chosen by indexing the coefficient table based on TarInc. It is important to note that Table A is only an example of an implementation. This table could be extended to support any size TarInc and generate any Filter size and Filter factor.

Once TarInc is mapped, the filter size is held in a register 42 and fed to the left and right kernel multiplier-accumulators 44 and 45. FIG. 5 illustrates this step. The fifth line of FIG. 5 (TarInc.RTM. Filter Mapping) shows the TarInc values from the fourth line (Target Increment) mapped to various FIR filters based on the mapping of Table A, above. This is represented schematically in FIG. 5 by placing the size of the respective half kernel of the assymetrical filter on each side of the target pixel, indicating respective half kernels for that pixel. In FIG. 5, the third line (Target) shows the first target pixel being generated by a 9 Tap filter, the second target pixel being generated by an assymetrical filter being one-half of a 9 Tap filter and one-half of a 3 Tap filter, the third target pixel being generated by a 3 Tap filter, and the fourth target pixel being generated by an assymetrical filter being one-half of a 3 Tap filter and one-half of a 5 Tap filter.

As discussed above, TarInc is mapped to an FIR filter size according to the mapping Table A, via filter look-up table 41, the output of which (i.e. filter factor value) is stored successively in register 42. The left kernel MAC 44 (i.e. multiplier accumulator) reads the filter factor value stored in register 42 and generates half of the target pixel value and a first contribution to the center tap product value using the left half of the designated filter kernel and the center tap for the left-side half kernel. On the same clock cycle, the right kernel MAC 45 generates the other half of the target pixel value and a second contribution to the center tap product value using the right half of the filter kernel and the center tap for right-side half kernel. The right kernel MAC 45 then sums the first and second contributions to the center tap multiplication product and divides that summed value by two for obtaining an averaged center tap multiplication product, and finally sums the averaged center tap multiplication product with the left half of the target pixel value received from left kernel MAC 44 and the right half of the target pixel value, resulting in the final warped output pixel.

As an alternative to summing the first and second contributions to the center tap multiplication product and dividing that summed value by two, the center tap coefficient for each half kernel may simply be halved prior to storage in the associated coefficient ROM of left and right kernel MACs 44 and 45, or via the appropriate code where the invention is implemented in software.

The upsample increment generator 46 receives and right-shifts the TarInc value by a predetermined number of bits equivalent to the received filter factor to produce an upsample increment value. For the example of FIG. 5, the first target increment is 2.25, resulting in a filter factor of two which corresponds to a "divide-by-four" operation (i.e. right-shift TarInc by two bits resulting in an upsample increment of 0.5625). The upsample increment value (i.e. right-shifted TarInc value) represents the correct equal spacing of upsampled pixels for application to the required size of FIR filter (e.g. 9-Tap filter for TarInc=2.25).

Interpolator 47 reads the upsample increment value and interpolates upsampled pixels from the source pixels on a continuous basis and with the spacing defined by the upsample increment value, until an internal incrementor or counter value reaches TarInc. For the example discussed above and shown in FIG. 5, four upsampled pixels are generated with the spacing of 0.5625 relative to the source pixel spacing, for application to the 9 Tap FIR filter. The interpolator 3 may be implemented as a well known linear interpolator, a quadratic interpolator as disclosed in applicant's co-pending application Ser. No. 08/172,065, or any other suitable design of interpolator.

The upsampled pixels generated by interpolator 47 are transmitted to the left and right kernel MACs 44 and 45 which contain control logic to ensure that the correct coefficient is multiplied by the correct upsampled pixel. The FIR filters shown operating between lines 2 and 3 of FIG. 5 are implemented by left and right half kernel MACs 44 and 45 which multiply the correct coefficients by the correct upsampled pixel values.

The control logic of left and right kernel MACs 44 and 45 also generates the first target pixel on start up of the system by doubling the right half kernel values for the first target pixel, in the manner disclosed in application Ser. No. 08/126,388. The last target pixel is generated by doubling the left kernel values. This doubling is required to ensure that all FIR filters maintain unity gain.

Other embodiments and variations of the invention are possible. For example, according to a further alternative embodiment, the invention is implemented in software using C programming language. This alternative software embodiment is fully disclosed in Appendix "A" to this disclosure. Although the warping performed in accordance with the preferred embodiment has been described as providing proper Nyquist bandlimiting, special effect warping can also be implemented using the principles of the present invention without Nyquist bandlimiting (eg. soft focus warping). In other words, the pre-filtering and interpolations steps can be implemented without Nyquist band-limiting. All such alternative embodiments and variations are believed to be within the sphere and scope of the claims appended hereto. ##SPC1##

Claims

1. A sample sequence warping system for receiving and warping an input sequence of source samples to produce an output sequence of target samples, comprising:

a) first means for receiving successive target increments representing desired target spacing between respective ones of said target samples, and in response generating successive filter factor values representing respective FIR filters having sufficient filter coefficients to generate respective ones of said target samples with said desired target spacing;

b) second means for receiving said successive filter factor values and in response selecting predetermined left and right half kernels of said respective FIR filters for generating said respective ones of said target samples;

c) third means for receiving said successive target increments and said successive filter factor values and in response generating a predetermined number of upsampled intermediate samples from said input sequence of source samples; and

d) fourth means for applying said upsampled intermediate samples to said predetermined left and right half kernels of said respective FIR filters for generating said output sequence of target samples.

2. The sample sequence warping system of claim 1, wherein said first means comprises a filter look-up table for correlating each said target increment with a respective one of said filter factor values and sizes of said FIR filters.

3. The sample sequence warping system of claim 2, wherein said filter look-up table generates a filter factor value of 0 when said target increment is between 0 and 1.

4. The sample sequence warping system of claim 2, wherein said filter look-up table generates a filter factor value of 1 when said target increment is between 1 and 2.

5. The sample sequence warping system of claim 2, wherein said filter look-up table generates a filter factor value of 2 when said target increment is between 2 and 4.

6. The sample sequence warping system of claim 2, wherein said filter look-up table generates a filter factor value of 3 when said target increment is between 4 and 8.

7. The sample sequence warping system of claim 2, wherein said filter look-up table generates a filter factor value of 4 when said target increment is between 8 and 16.

8. The sample sequence warping system of claim 2, wherein said filter look-up table generates a filter factor value of 5 when said target increment is between 16 and 32.

9. The sample sequence warping system of claim 1, wherein said second means further comprises means for selecting said predetermined left and right half kernels of the largest one of said respective FIR filters on either side of each of said target samples.

10. The sample sequence warping system of claim 9, wherein said means for selecting said predetermined left and right half kernels further comprises:

i) first register means connected to said first means, for receiving and delaying said successive filter factor values by one sample period; and

ii) first comparator means connected to said first means and said first register means, for selecting the largest of two successive filter factor values output from said first means and said first register means, and in response outputing successive largest ones of two said successive filter factor values.

11. The sample sequence warping system of claim 10, wherein said third means further comprises:

iii) second register means connected to said first register means for receiving and delaying said successive filter factor values by a further sample period;

iv) second comparator means connected to said first means, said first register means and said second register means, for selecting the largest of three successive filter factor values output from respective ones of said first means, said first register means and said second register means, and in response outputing successive largest ones of three said successive filter factor values;

v) an upsample increment generator for receiving said successive target increments and said successive largest ones of three said successive filter factor values and in response right shifting said successive target increments respectively by said successive largest ones of three said filter factor values to form successive upsample increment values; and

vi) interpolator means for receiving said successive binary target increments, said successive upsample increment values and said input sequence of source samples, and in response interpolating said input sequence of source samples to produce said predetermined number of upsampled intermediate samples spaced apart by a respective one of said upsample increment values, said predetermined number being equivalent to a respective one of said target increments divided by said respective one of said upsample increment values.

12. The sample sequence warping system of claim 11, wherein said fourth means further comprises:

vii) third register means connected to said second comparator means for receiving and delaying said successive largest ones of two said successive filter factor values by one sample period;

viii) fourth register means connected to said third register means for receiving and delaying said successive largest ones of two said successive filter factor values by a further sample period;

ix) first half digital filter kernel multiplier accumulator means connected to said third register means and said interpolator means, for receiving and multiplying a first plurality of said upsampled intermediate samples by successive ones of said filter coefficients identified by said successive largest ones of two said successive filter factor values delayed by one sample period and in response generating a first plurality of intermediate product values, summing said first plurality of intermediate product values and in response generating a succession of accumulated intermediate product values; and

x) second half digital filter kernel multiplier accumulator means connected to said first half digital filter kernel multiplier accumulator means, said fourth register means and said interpolator means, for receiving and multiplying a second plurality of said successive upsampled intermediate samples by successive ones of said filter coefficients identified by said successive largest ones of two said successive filter factor values delayed by a further sample period and in response generating a second plurality of intermediate product values, summing said second plurality of intermediate product values and in response generating a further succession of accumulated intermediate product values, and summing said succession of accumulated intermediate product values and said further succession of accumulated intermediate product values to generate said output sequence of target samples.

13. The sample sequence warping system of claim 1, wherein said second means further comprises means for selecting said predetermined left and right half kernels from respective ones of said FIR filters on either side of each of said target samples.

14. The sample sequence warping system of claim 13, wherein said third means further comprises:

i) an upsample increment generator for receiving said successive target increments and said successive filter factor values and in response right shifting said successive target increments respectively by said successive filter factor values to form successive upsample increment values; and

ii) interpolator means for receiving said successive target increments, said successive upsample increment values and said input sequence of source samples, and in response interpolating said input sequence of source samples to produce said predetermined number of upsampled intermediate samples spaced apart by a respective one of said upsample increment values, said predetermined number being equivalent to a respective one of said target increments divided by said respective one of said upsample increment values.

15. The sample sequence warping system of claim 14, wherein said fourth means further comprises:

iii) register means connected to said first means, for receiving and delaying said successive filter factor values by one sample period;

iv) first half digital filter kernel multiplier accumulator means connected to said register means and said interpolator means, for receiving and multiplying a first plurality of said upsampled intermediate samples by successive ones of said filter coefficients identified by said successive filter factor values delayed by one sample period and in response generating a first plurality of intermediate product values, summing said first plurality of intermediate product values and in response generating a succession of accumulated intermediate product values; and

v) second half digital filter kernel multiplier accumulator means connected to said first half digital filter kernel multiplier accumulator means, said register means and said interpolator means, for receiving and multiplying a second plurality of said successive upsampled intermediate samples by successive ones of said filter coefficients identified by said successive filter factor values delayed by said one sample period and in response generating a second plurality of intermediate product values, summing said second plurality of intermediate product values and in response generating a further succession of accumulated intermediate product values, and summing said succession of accumulated intermediate product values and said further succession of accumulated intermediate product values to generate said output sequence of target samples.

16. A sample sequence warping method for receiving and warping an input sequence of source samples to produce an output sequence of target samples, comprising the steps of:

a) receiving successive target increments representing desired target spacing between respective ones of said target samples, and in response generating successive filter factor values representing respective FIR filters having sufficient filter coefficients to generate respective ones of said target samples with said desired target spacing;

b) receiving said successive filter factor values and in response selecting predetermined left and right half kernels of said respective FIR filters for generating said respective ones of said target samples;

c) receiving said successive target increments and said successive filter factor values and in response generating a predetermined number of upsampled intermediate samples from said input sequence of source samples; and

d) applying said upsampled intermediate samples to said predetermined left and right half kernels of said respective FIR filters for generating said output sequence of target samples.

17. The sample sequence warping method of claim 16, wherein said step of receiving said successive filter factor values and in response selecting predetermined left and right half kernels of said respective FIR filters for generating said respective ones of said target samples further comprises selecting said predetermined left and right half kernels of the largest one of said respective FIR filters on either side of each of said target samples.

18. The sample sequence warping method of claim 17, wherein said step of selecting said predetermined left and right half kernels further comprises:

i) receiving and delaying said successive filter factor values by one sample period; and

ii) selecting the largest of two successive filter factor values and in response outputing successive largest ones of two said successive filter factor values.

19. The sample sequence warping method of claim 18, wherein said step of receiving said successive target increments and said successive filter factor values and in response generating a predetermined number of upsampled intermediate samples from said input sequence of source samples further comprises:

iii) receiving and delaying said successive filter factor values by a further sample period;

iv) selecting the largest of three successive filter factor values and in response outputing successive largest ones of three said successive filter factor values;

v) receiving said successive target increments and said successive largest ones of three said successive filter factor values and in response right shifting said successive target increments respectively by said successive largest ones of three said filter factor values to form successive upsample increment values; and

vi) receiving said successive binary target increments, said successive upsample increment values and said input sequence of source samples, and in response interpolating said input sequence of source samples to produce said predetermined number of upsampled intermediate samples spaced apart by a respective one of said upsample increment values, said predetermined number being equivalent to a respective one of said target increments divided by said respective one of said upsample increment values.

20. The sample sequence warping method of claim 19, wherein said step of applying said upsampled intermediate samples to said predetermined left and right half kernels of said respective FIR filters for generating said output sequence of target samples further comprises:

vii) receiving and delaying said successive largest ones of two said successive filter factor values by one sample period;

viii) receiving and delaying said successive largest ones of two said successive filter factor values by a further sample period;

ix) receiving and multiplying a first plurality of said upsampled intermediate samples by successive ones of said filter coefficients identified by said successive largest ones of two said successive filter factor values delayed by one sample period and in response generating a first plurality of intermediate product values, summing said first plurality of intermediate product values and in response generating a succession of accumulated intermediate product values; and

x) receiving and multiplying a second plurality of said successive upsampled intermediate samples by successive ones of said filter coefficients identified by said successive largest ones of two said successive filter factor values delayed by a further sample period and in response generating a second plurality of intermediate product values, summing said second plurality of intermediate product values and in response generating a further succession of accumulated intermediate product values, and summing said succession of accumulated intermediate product values and said further succession of accumulated intermediate product values to generate said output sequence of target samples.

21. The sample sequence warping method of claim 16, wherein said step of receiving said successive filter factor values and in response selecting predetermined left and right half kernels of said respective FIR filters for generating said respective ones of said target samples further comprises selecting said predetermined left and right half kernels from respective ones of said FIR filters on either side of each of said target samples.

22. The sample sequence warping method of claim 21, wherein said step of receiving said successive target increments and said successive filter factor values and in response generating a predetermined number of upsampled intermediate samples from said input sequence of source samples further comprises:

i) receiving said successive target increments and said successive filter factor values and in response right shifting said successive target increments respectively by said successive filter factor values to form successive upsample increment values; and

ii) receiving said successive target increments, said successive upsample increment values and said input sequence of source samples, and in response interpolating said input sequence of source samples to produce said predetermined number of upsampled intermediate samples spaced apart by a respective one of said upsample increment values, said predetermined number being equivalent to a respective one of said target increments divided by said respective one of said upsample increment values.

23. The sample sequence warping method of claim 22, wherein said step of applying said upsampled intermediate samples to said predetermined left and right half kernels of said respective FIR filters for generating said output sequence of target samples further comprises:

iii) receiving and delaying said successive filter factor values by one sample period;

iv) receiving and multiplying a first plurality of said upsampled intermediate samples by successive ones of said filter coefficients identified by said successive filter factor values delayed by one sample period and in response generating a first plurality of intermediate product values, summing said first plurality of intermediate product values and in response generating a succession of accumulated intermediate product values; and

v) receiving and multiplying a second plurality of said successive upsampled intermediate samples by successive ones of said filter coefficients identified by said successive filter factor values delayed by said one sample period and in response generating a second plurality of intermediate product values, summing said second plurality of intermediate product values and in response generating a further succession of accumulated intermediate product values, and summing said succession of accumulated intermediate product values and said further succession of accumulated intermediate product values to generate said output sequence of target samples.