Approximations used in performance sensitive transformations which contain sub-transforms
View Patent ↗The present invention provides methods, apparatus and articles of manufacture for performing a transform equation in which at least two sub-transform equations can be identified. According to the invention, each sub-transform is considered separately and the constants of each sub-transform are replaced with an approximation which is a function of an integer value and a second value, where the second value is common to each constant in a sub-transform. The transform equation is then performed with the integer values in place of the constants. As a result either the input to the transform equation or at least one of the result of the transform and a value to be compared the result, is modified with a value which is a function of the second values for each sub-transform. By considering each sub-transform separately the second values of each sub-transform can be different thus allowing more accurate approximations to be used with minimal degradation in performance.
1. A method of reducing an error in a result of a transform equation, the method comprising:
identifying at least two sub-transforms in the transform equation;
approximating the constants in each of the at least two sub-transforms using denominators and integer numerators, wherein the denominators are different for each of the at least two sub-transform constants;
replacing the constants in each of the at least two sub-transforms with the integer numerators of the approximated constants to generate at least two modified sub-transforms;
generating the result by performing the transform equation using the at least two modified sub-transforms; and
modifying the result using the denominator of each of the at least two sub-transforms to reduce the error.
2. The method of claim 1 wherein the denominators are floating point values.
3. The method of claim 2 wherein the denominators are fixed precision values.
4. The method of claim 1 wherein modifying the result is performed as part of a quantization step.
5. A method of reducing an error in a result of a transform equation, the method comprising:
modifying input data for the transform;
identifying at least two sub-transforms in the transform equation;
approximating the constants in each of the at least two sub-transforms using denominators and integer numerators, wherein the denominators are different for each of the at least two sub-transform constants;
replacing the constants in each of the at least two sub-transforms with the integer numerators of the approximated constants to generate at least two modified sub-transforms;
generating the result by performing the transform equation using the at least two modified sub-transforms; and
modifying the result using the denominator of each of the at least two sub-transforms to reduce the error.
6. The method of claim 5 wherein the denominators are floating point values.
7. The method of claim 6 wherein the denominators are fixed precision values.
8. The method of claim 5 wherein modifying the result is part of a de-quantization step.
9. A data analyzing system comprising:
a transformer which performs a selected transform equation having a plurality of sub-transform equations, using pre-selected integer values in place of the transform constants, to produce a result; and
a modifier which modifies at least one of the result of the transform equation and a comparison value to be compared with the result, using a pre-selected value, wherein each constant of a sub-transform approximates to a function of the pre-selected integer used in its place and a second value which is common to each transform constant in the sub-transform, and wherein the pre-selected value is a function of second values associated with each sub-transform.
10. The data analyzing system of claim 9 wherein the second values are floating point values.
11. The data analyzing system of claim 10 wherein the pre-selected value is a fixed precision approximation of a value obtained from the function of the second values.
12. A data analyzing system comprising:
a modifier which modifies input data for a selected transform equation comprising at least two sub-transforms using a pre-selected value, and
a transformer which performs the selected transform equation on the modified input data using pre-selected integer values in place of the transform constants, to produce a result, wherein each constant of a sub-transform approximates to a function of the pre-selected integer used in its place and a second value which is common to each transform constant in the sub-transform, and wherein the pre-selected value is a function of the second values for each sub-transform.
13. The data analyzing system of claim 12 wherein the second values are floating point values.
14. The data analyzing system of claim 13 wherein the pre-selected value is a fixed precision approximation of a value obtained from the function of the second values.
15. An apparatus comprising:
memory for storing image data;
a processor for processing the image data to provide a compressed print stream output; and
a printhead driving circuit for controlling a printhead to generate a printout of the image data;
wherein the processor applies a pre-selected transform equation comprising at least two sub-transforms to the image data by:
modifying the image using a pre-selected value, and
performing the selected transform equation on the modified image data using pre-selected integer values in place of the transform constants, to produce a result, wherein each constant of a sub-transform approximates to a function of the pre-selected integer used in its place and a second value which is common to each transform constant in the sub-transform, and wherein the pre-selected value is a function of the second values for each sub-transform.
16. The apparatus of claim 15 wherein the second values are floating point values.
17. The apparatus of claim 16 wherein the pre-selected value is a fixed precision approximation of a value obtained from the function of the second values.
18. An article of manufacture comprising a program storage medium readable by a computer, the medium tangibly embodying one or more programs of instructions executable by a computer to perform a method of reducing an error in a result of a transform equation, the method comprising:
identifying at least two sub-transforms in the transform equation;
approximating the constants in each of the at least two sub-transforms using denominators and integer numerators, wherein the denominators are different for each of the at least two sub-transform constants;
replacing the constants in each of the at least two sub-transforms with the integer numerators of the approximated constants to generate at least two modified sub-transforms;
generating the result by performing the transform equation using the at least two modified sub-transforms; and
modifying the result using the denominator of each of the at least two sub-transforms to reduce the error.
19. The article of manufacture of claim 18 wherein the denominators are floating point values.
20. The article of manufacture of claim 19 wherein the denominators are fixed precision values.
21. The article of manufacture of claim 18 wherein modifying the result is performed as part of a quantization step.
22. An article of manufacture comprising a program storage medium readable by a computer, the medium tangibly embodying one or more programs of instructions executable by a computer to perform a method to reduce an error in a result of a transform equation, the method comprising:
modifying input data for the transform;
identifying at least two sub-transforms in the transform equation;
approximating the constants in each of the at least two sub-transforms using denominators and integer numerators, wherein the denominators are different for each of the at least two sub-transform constants;
replacing the constants in each of the at least two sub-transforms with the integer numerators of the approximated constants to generate at least two modified sub-transforms;
generating the result by performing the transform equation using the at least two modified sub-transforms; and
modifying the result using the denominator of each of the at least two sub-transforms to reduce the error.
23. The article of manufacture of claim 22 wherein the denominators are floating point values.
24. The article of manufacture of claim 23 wherein the denominators are fixed precision values.
25. The article of manufacture of claim 22 wherein modifying the result is part of a de-quantization step.