Signal-charge estimator and method

Patent metadata

Field	Value
Publication number	US-11871129-B2
Application number	US-202217869552-A
Country	US
Kind code	B2
Filing date	Jul 20, 2022
Priority date	Jan 11, 2022
Publication date	Jan 9, 2024
Grant date	Jan 9, 2024

Abstract

A method for estimating a signal charge collected by a pixel of an image sensor includes determining an average bias that depends on the pixel's floating-diffusion dark current and pixel-sampling period. The method also includes determining a signal-charge estimate as the average bias subtracted from a difference between a weighted sum of a plurality of N multiple-sampling values each multiplied by a respective one of a plurality of N sample-weights.

First claim

What is claimed is:

1. A method for estimating a signal charge collected by a pixel of an image sensor, comprising: determining an average bias that depends on the pixel's floating-diffusion dark current and pixel-sampling period; determining a plurality of N sample-weights, as a plurality of N optimal sample-weights by solving a system of linear equations that are expressible as a product of (i) a matrix that includes a covariance matrix of a plurality of N multiple-sampling values and (ii) a vector that includes the plurality of N optimal sample-weights, wherein N is a positive integer greater than one; and determining a signal-charge estimate as the average bias subtracted from a difference between a weighted sum of the plurality of N multiple-sampling values each multiplied by a respective one of the plurality of N sample-weights that (i) collectively minimize an expected value of the signal-charge estimate squared and (ii) sum to unity.

2. The method of claim 1 , wherein solving the system of linear equations includes inverting the matrix.

3. A method for estimating a signal charge collected by a pixel of an image sensor, comprising: determining an average bias that depends on the pixel's floating-diffusion dark current and pixel-sampling period; determining a signal-charge estimate as the average bias subtracted from a difference between a weighted sum of a plurality of N multiple-sampling values each multiplied by a respective one of a plurality of N sample-weights that (i) collectively minimize an expected value of the signal-charge estimate squared and (ii) sum to unity wherein N is a positive integer greater than one, each of the plurality of multiple-sampling values being proportional to a difference between (i) a respective one of a plurality of N reset-samples output by the pixel and (ii) a respective one of a plurality of N signal-samples output by the pixel, each of the plurality of N reset-samples including a read-noise component, and further comprising: determining the read-noise component as a weighted sum of Lorentzian noise sources each having a respective time-constant.

4. The method of claim 3 , each term of the weighted sum of Lorentzian noise sources including a respective noise-weight of a noise-weight set, and further comprising:
determining values of (i) each of the respective noise-weights, (ii) each of the respective time-constants, and (iii) the floating-diffusion dark current that minimize a sum of a plurality of differences between (a) the sample variance of the plurality of N multiple-sampling values and (b) a model-based sample-variance that depends on the read-noise component, each of the plurality of differences corresponding to a respective time delay between samples.

5. The method of claim 4 , said determining values comprising, before determining each of the respective noise-weights and the floating-diffusion dark current:
determining each of the respective time-constants independently of the sum of the plurality of differences.

6. The method of claim 4 , further comprising determining the plurality of N sample-weights, as a plurality of N optimal sample-weights, by:
determining, for each pair of multiple-sampling values of the plurality of N multiple-sampling values, a respective covariance of the pair of multiple-sampling values that depends on the noise-weight set and the floating-diffusion dark current; and
inverting a matrix that includes a matrix of the covariances, wherein the plurality of N optimal sample-weights (i) collectively minimize an expected value of the signal-charge estimate squared and (ii) sum to unity.

7. The method of claim 4 , further comprising:
for each of a plurality of additional pixels of the image sensor, repeating the method of claim 6 to yield a plurality of additional noise-weight sets and a plurality of additional floating-diffusion dark currents;
averaging each of the noise-weight set and the plurality of additional noise-weight sets to yield an average noise-weight set; and
averaging the floating-diffusion dark current and the plurality of additional floating-diffusion dark currents to yield an average floating-diffusion dark current and a mean-square floating-diffusion dark current.

8. The method of claim 7 , further comprising determining the plurality of N sample-weights, as a plurality of N optimal sample-weights, by:
determining, for each pair of multiple-sampling values of the plurality of N multiple-sampling values, a covariance of the pair of multiple-sampling values that depends on the average noise-weight set, the average floating-diffusion dark current. and the mean-square floating-diffusion dark current; and
inverting a matrix that includes a matrix of the covariances, wherein the plurality of N optimal sample-weights (i) collectively minimize an expected value of the signal-charge estimate squared and (ii) sum to unity.

9. The method of claim 4 , further comprising:
determining the average bias from the floating-diffusion dark current and the pixel-sampling period.

10. The method of claim 3 , further comprising determining the weighted sum of the plurality of N multiple-sampling values by:
reading a first reset-sample of the plurality of N reset-samples from an analog-to-digital converter of the image sensor;
storing, in a memory, a cumulative weighted-reset-sample equal to a product of (i) the first reset-sample and (ii) a first sample-weight of the plurality of N sample-weights;
for each reset-sample and sample-weight of the plurality of N reset-samples and N sample-weights other than the first reset-sample and the first sample-weight, (i) reading the reset-sample from the analog-to-digital converter, and (ii) increasing the cumulative weighted-reset-sample by a product of the reset-sample and the sample-weight;
reading a first signal-sample of the plurality of N reset-samples from the analog-to-digital converter;
storing, in a memory, a cumulative weighted-signal-sample equal to a product of (i) the first signal-sample and (ii) the first sample-weight of the plurality of N sample-weights;
for each signal-sample and sample-weight of the plurality of N signal-samples and N sample-weights other than the first signal-sample and first sample-weight, (i) reading the signal-sample from the analog-to-digital converter, and (ii) increasing the cumulative weighted-signal-sample by a product of the signal-sample and the sample-weight; and
determining the weighted sum as a quotient of (i) a difference between the cumulative weighted-reset-sample and the cumulative weighted-signal-sample and (ii) a product of N and a conversion gain of the pixel.

11. The method of claim 3 , further comprising determining the weighted sum of the plurality of N multiple-sampling values by:
reading a first reset-sample of the plurality of N reset-samples from an analog-to-digital converter of the image sensor;
reading a first signal-sample of the plurality of N reset-samples from the analog-to-digital converter;
storing, in a memory, a cumulative weighted-sample equal to a product of (i) a difference between the first reset-sample and the first signal-sample, and (ii) a first sample-weight of the plurality of N sample-weights;
for each of the plurality of N reset-samples, N signal-samples, and N sample-weights other than the first reset-sample, the first signal-sample and the first sample-weight, (i) reading the reset-sample and the signal-sample from the analog-to-digital converter, and (ii) increasing the cumulative weighted-sample by a product of (a) a difference between the reset-sample and the signal-sample and (b) the sample-weight; and
determining the weighted sum as a quotient of (i) the cumulative weighted-sample

Assignees

• Omnivision Tech Inc

Inventors

• Fowler Boyd
• Suess Andreas

Classifications

• H04N25/772

  comprising A/D, V/T, V/F, I/T or I/F converters · CPC title

• H04N25/671 Primary

  for non-uniformity detection or correction · CPC title

• G06F17/12

  Simultaneous equations {, e.g. systems of linear equations} · CPC title

• H04N25/62 Primary

  Detection or reduction of noise due to excess charges produced by the exposure, e.g. smear, blooming, ghost image, crosstalk or leakage between pixels · CPC title

• H04N25/63

  applied to dark current · CPC title