Correcting low-resolution measurements

1 . A method comprising:

measuring, by a low-resolution measurement device including at least one sensor, a plurality of low-resolution measurements, the plurality of low-resolution measurements being quantitative measurements that are rounded versions of a corresponding plurality of unobservable high-resolution measurements and therefore having a lower resolution than the resolution of the corresponding plurality of unobservable high-resolution measurements;

introducing variation to the plurality of low-resolution measurements by iteratively computing, until a termination criteria is met, corresponding perturbed measurements for the low-resolution measurements, said corresponding perturbed measurements having a higher resolution than another resolution of the low-resolution measurements, a first final set of corresponding perturbed measurements being obtained after the termination criteria is met;

performing the introducing a predefined number of times to obtain other final sets of corresponding perturbed measurements;

identifying the final set of corresponding perturbed measurements having a median shape parameter;

running, a distribution test for a Weibull distribution on the identified final set of corresponding perturbed measurements.

2 . The method of claim 1 , wherein the distribution test is an Anderson-Darling test for the Weibull distribution.

3 . The method of claim 1 , wherein the plurality of low-resolution measurements are determined to be from the Weibull distribution.

4 . The method of claim 1 , wherein the introducing comprises:

computing, for each low-resolution measurement, a first interval that contains a corresponding unobservable high-resolution measurement corresponding to said each low-resolution measurement;

generating, for each low-resolution measurement, a random observation from a uniform distribution on a defined interval;

estimating distribution parameters of said low-resolution measurements by applying a continuous-data Weibull parameter estimation to the plurality of low-resolution measurements;

transforming each random observation to be uniform on a second interval that corresponds to a distribution function of the first interval to obtain corresponding rescaled uniform observations, said distribution function being based on estimated distribution parameters of said low-resolution measurements; and

inverse transforming, responsive to the transforming, and using an inverse of the distribution function, said rescaled uniform observations to obtain said corresponding perturbed values,

wherein the transforming and inverse transforming are repeated iteratively using new estimated distribution parameters of the corresponding perturbed values until said termination criteria is met.

5 . The method of claim 4 , wherein the estimated distribution parameters are a shape and a scale parameter.

6 . The method of claim 4 , further comprising:

replacing, in the estimating, low resolution measurements having a value of zero with a positive value.

7 . The method of claim 6 , the positive value is a midpoint of the interval of positive unobservable high-resolution measurements that are rounded to zero by the low-resolution measurement device.

8 . The method of claim 1 , wherein low-resolution measurement device is a device selected from the list consisting of a medical device, a vision/imaging device, a detector, a transducer, and an instrument used in measuring physical quantities and attributes of real-world objects and events.

9 . The method of claim 1 , wherein the termination criteria is a stable shape parameter condition.

10 . The method of claim 1 , further comprising:

obtaining the plurality of low-resolution measurements by measuring values of a defined property, using a low-resolution measurement device and wherein the values of the defined property are quantitative values.

11 . A method comprising:

measuring, by a low-resolution measurement device including at least one sensor, a plurality of low-resolution measurements, the plurality of low-resolution measurements being quantitative measurements that are rounded versions of a corresponding plurality of unobservable high-resolution measurements and therefore having a lower resolution relative to the resolution of the corresponding plurality of unobservable high-resolution measurements;

introducing variation to the plurality of low-resolution measurements by computing corresponding perturbed measurements for the low-resolution measurements, said corresponding perturbed measurements having a higher resolution than another resolution of the low-resolution measurements;

running, a distribution test for a Weibull distribution on the corresponding perturbed measurements.

12 . The method of claim 11 , wherein the distribution test is an Anderson-Darling test for the Weibull distribution.

13 . The method of claim 11 , wherein the corresponding perturbed measurements are computed based on computing parameter estimates that minimize a Kullback-Leibler divergence representing the statistical distance of a discretized Weibull distribution from observed frequencies of the low-resolution measurements.

14 . The method of claim 13 , further comprising:

computing, for each low-resolution measurement, a first interval that contains a corresponding unobservable high-resolution measurement corresponding to said each low-resolution measurement;

generating, for each low-resolution measurement, a random observation from a uniform distribution on a defined interval;

transforming each random observation to be uniform on a second interval that corresponds to a distribution function of the first interval to obtain corresponding rescaled uniform observations, said distribution function being based on the parameter estimates; and

inverse transforming, responsive to the transforming, and using an inverse of the distribution function, said rescaled uniform observations to obtain said corresponding perturbed measurements.

15 . The method of claim 13 , the parameter estimates are a shape and a scale parameter.

16 . The method of claim 11 , wherein the low-resolution measurements are rounded versions of the unobservable high-resolution measurements.

17 . The method of claim 11 wherein the distribution test is an Anderson-Darling test for the Weibull distribution.

18 . A non-transitory computer readable storage medium storing program instructions which, when executed by a processor, causes the processor to perform a procedure comprising the steps of:

measuring, by a low-resolution measurement device including at least one sensor, a plurality of low-resolution measurements, the plurality of low-resolution measurements being quantitative measurements that are rounded versions of a corresponding plurality of unobservable high-resolution measurements and therefore having a lower resolution than the resolution of the corresponding plurality of unobservable high-resolution measurements;

introducing variation to the plurality of low-resolution measurements by iteratively computing, until a termination criteria is met, corresponding perturbed measurements for the low-resolution measurements, said corresponding perturbed measurements having a higher resolution than another resolution of the low-resolution measurements, a first final set of corresponding perturbed measurements being obtained after the termination criteria is met;

performing the introducing a predefined number of times to obtain other final sets of corresponding perturbed measurements;

identifying the final set of corresponding perturbed measurements having a median shape parameter;

running, a distribution test for a Weibull distribution on the identified final set of corresponding perturbed measurements.

19 . A computer system comprising:

a low-resolution measurement device including at least one sensor, configured to measure a plurality of low-resolution measurements, the plurality of low-resolution measurements being quantitative measurements that are rounded versions of a plurality of unobservable high-resolution measurements and therefore having a lower resolution that the resolution of the corresponding plurality of unobservable high-resolution measurements; and

at least one processor configured to performs the steps of:

introducing variation to the plurality of low-resolution measurements by iteratively computing, until a termination criteria is met, corresponding perturbed measurements for the low-resolution measurements, said corresponding perturbed measurements having a higher resolution than another resolution of the low-resolution measurements, a first final set of corresponding perturbed measurements being obtained after the termination criteria is met;

performing the introducing and running a predefined number of times to obtain other final sets of corresponding perturbed measurements;

identifying the final set of corresponding perturbed measurements having a median shape parameter;

running, a distribution test for a Weibull distribution on the identified final set of corresponding perturbed measurements.

20 . A non-transitory computer readable storage medium storing program instructions which, when executed by a processor, causes the processor to perform a procedure comprising the steps of:

measuring, by a low-resolution measurement device including at least one sensor, a plurality of low-resolution measurements, the plurality of low-resolution measurements being quantitative measurements that are rounded versions of a corresponding plurality of unobservable high-resolution measurements and therefore having a lower resolution that the resolution of the corresponding plurality of unobservable high-resolution measurements;

introducing variation to the plurality of low-resolution measurements by computing corresponding perturbed measurements for the low-resolution measurements, said corresponding perturbed measurements having a higher resolution than another resolution of the low-resolution measurements;

running, a distribution test for a Weibull distribution on the corresponding perturbed measurements.

21 . A computer system comprising:

a low-resolution measurement device including at least one sensor, configured to measure a plurality of low-resolution measurements, the plurality of low-resolution measurements being quantitative measurements that are rounded versions of a corresponding plurality of unobservable high-resolution measurements and therefore having a lower resolution that the resolution of the corresponding plurality of unobservable high-resolution measurements; and

at least one processor configured to performs the steps of:

introducing variation to the plurality of low-resolution measurements by computing corresponding perturbed measurements for the low-resolution measurements, said corresponding perturbed measurements having a higher resolution than another resolution of the low-resolution measurements;

running, a distribution test for a Weibull distribution on the corresponding perturbed measurements.