Patent history

An early formulation of the DEBRA algorithm was patented. The patent expired in 2024; this page records the public record. Development of this implementation began in August 2026, more than two years after the patent lapsed.

Not legal advice

This page is historical documentation of public patent records, last verified 2026-08-29 against the sources linked at the bottom. It is not legal advice.

The patent

Grant

US 9,383,478 B2, “System and method for atmospheric parameter enhancement”

Inventor

Steven D. Miller (the paper’s first author)

Assignee

The USA as represented by the Secretary of the Navy (reel/frame 037279/0473, executed 2015-12-02)

Application

14/150,467, filed 2014-01-08

Priority

Provisional 61/756,555, 2013-01-25

Granted

2016-07-05

Nominal expiration

2034-09-01

Actual status

Expired 2024-07-05 for non-payment of the 8th-year maintenance fee (37 CFR 1.362)

The USPTO legal-event record: the 4th-year maintenance fee was paid (2019-12-18), the 8th-year reminder was mailed 2024-02-26, no payment followed, the lapse was recorded 2024-08-12 with effect from 2024-07-05. No petition to reinstate appears in the Google Patents legal-event record as of the verification date. Whether one is pending in the USPTO’s own transaction history has not been checked; Patent Center requires a login.

No patent outside the United States

The patent’s only family member (DOCDB family 51223030) is PCT application WO 2014/116472 A1 (PCT/US2014/011586), which ceased without entering the national phase in any country. There has never been a corresponding patent in Japan, China, Europe, or anywhere else. There are also no continuations, divisionals, or continuations-in-part within this family: application 14/150,467 is the only US member of DOCDB family 51223030.

Note

That statement is scoped to the family. A later application by the same inventor — an ABI adaptation, or an ash or fog variant — would not appear in family 51223030 and is not covered here. Establishing that would take an inventor-name search of the USPTO full-text database, which has not been done.

Two predecessor patents by the same inventor and assignee — US 7,242,803 and US 7,379,592, the SeaWiFS-era “significant dust detection and enhancement” algorithms (priority 2003) — expired at the end of their 20-year terms.

Constants in the patent differ from the published algorithm

The patent recites specific constants for the algorithm’s tests. They predate the published paper and differ from what this package (which follows Miller et al. 2017, as amended by the 2020 erratum) implements. Recorded here as provenance for anyone cross-checking where a number came from:

Quantity

Patent claims (2014)

Miller et al. 2017 / this package

Reference window channel

11.2 µm

10.3/10.4 µm

Split-window (BTD1 / DT1) upper bound

4.0 K

3.5 K

8.5 µm test (BTD2 / DT2) upper bound

0.5 K

3.0 K

Confidence normalization

0.5 – 2.5

0.25 – 2.50

Terminator weighting

threshold on cos θ > 0.383

zenith-band blending, exponent 1.5

RGB composition: CF cap / blue dimming / gun max

0.5 / 0.1 / 1.2

0.5 / 0.1 / 1.2

Lineage, 2001–2017

Note

This section is a reconstruction assembled from public documents — the patents and papers cited here. The equations, channel tables, and dates are transcribed from those sources. The sequence, and the reasoning attributed to each change, are inferences drawn from reading them; they are not an account given by the authors. Nothing on this page has been reviewed or confirmed by Steven D. Miller, the Naval Research Laboratory, CIRA, or Colorado State University.

The algorithm predates the 2013 application. Its two predecessors — US 7,242,803 (application 10/713,908, recited on the US 7,379,592 front page as filed 2003-01-21) and its continuation-in-part US 7,379,592 (application 10/885,526, filed 2004-06-30) — describe a product that was already in routine operational use and had already been revised. Automated processing of the SeaWiFS data “commenced Aug. 8, 2001”; the dust case in Fig. 3 was captured at the Navy Regional Center in Rota, Spain, on 2001-02-13 1255Z, one of the three receiving stations named in the text along with Bahrain and Yokosuka.

Everything quoted or transcribed below is from the US 7,379,592 specification unless stated otherwise.

Four stages

Stage

When

What changed

Limitation of that stage

SeaWiFS, visible only

2001–2002

Dust over water read as a colour anomaly

No infrared channels; land not addressable

MODIS, “inclusive” logic (Eq. 3)

2003–2004

Thermal infrared added; the four terms summed

“the cost of this aggression is a high frequency of false alarms” — most commonly “cold land, cloud shadows, thin cirrus, and sunglint upon unmasked lake bodies”

“Exclusive” logic (Eq. 5)

2004

The same four terms multiplied, so all must agree

“it only takes a single zero-valued term to set the entire \(D_{\text{lnd}}^{\text{new}}\) term to zero”, and the split-window term weakens for very thick dust

DEBRA

2013–2017

Weighted maximum; background from external priors; masks become continuous weights; a per-pixel confidence factor as the output

—

The limitations quoted for the inclusive and exclusive stages are stated in the specification. Arranging the four into a single line of descent, each stage driven by the failures of the one before it, is an inference drawn from reading them in order; the sources do not present themselves that way.

The two 2004 formulations coexisted: “Equation 5 should not supplant Equation 3, but rather be considered as a variation of the technique providing superior results in certain dust scenarios.” As of 2004 the combination logic was, on the record, still unsettled.

The 2004 land algorithm, as recited

\(T(n)\) is the brightness temperature of MODIS channel \(n\) in kelvins, \(R(n)\) its normalized reflectance, and \(T_{\max}(31)\) the maximum pixel temperature in the current scene.

The red gun carries the dust enhancement; the blue and green guns are the same as in the over-water algorithm.

\[\begin{split} \begin{aligned} D_{\text{lnd}} &= L_1 + L_3 - L_4 + (1.0 - L_2) & &\text{(Eq. 3, inclusive; scaled } [1.3,\; 2.7]) \\ D_{\text{lnd}}^{\text{new}} &= L_1 \, (1.0 - L_2) \, L_3 \, (1.0 - L_4) & &\text{(Eq. 5, exclusive; scaled } [0.35,\; 0.75]) \end{aligned} \end{split}\]

The same four terms appear in both (Table 2):

Term

Expression

Normalization bounds

\(L_1\)

\(T(32) - T(31)\)

\(-2 \rightarrow 2\) K

\(L_2\)

\(T(31)\)

\(T_{\text{dyn}}(31) \rightarrow T_{\max}(31)\)

\(L_3\)

\(2R(1) - R(3) - R(4) - L_2\)

\(-1.5 \rightarrow 0.25\)

\(L_4\)

\(R(26) > 0.05\;?\;0\), else \(1\)

(n/a)

and the dynamic temperature floor that \(L_2\) scales against:

\[\begin{split} T_{\text{dyn}} = \begin{cases} T_{\max}(31) - 21 & \text{if } T_{\max}(31) < 301\;\text{K} \\[2pt] \bigl(T_{\max}(31) - 273\bigr) / 4 + 273 & \text{otherwise} \end{cases} \qquad \text{(Eq. 4)} \end{split}\]

Channels used (Table 1), with Rayleigh scatter removed from 1–4:

Channel

λ (µm)

Resolution (km)

Description

1

0.645

0.25

Red

2

0.853

0.25

Reflective IR

3

0.469

0.50

Blue

4

0.555

0.50

Green

26

1.38

1.0

Shortwave Vapor

31

11.0

1.0

IR Window 1

32

12.0

1.0

IR Window 2

The over-water branch of the same patent is a normalized difference of two SeaWiFS reflectances \(\alpha_\lambda\) at \(\lambda\) nm,

\[ \Delta = \frac{\alpha_{865} - \alpha_{412}}{\alpha_{865} + \alpha_{412}} \]

whose logarithm is “scaled between −0.45 and 0.20 and loaded into the red channel of the RGB composite”. Its cloud screen is a mean of the 412, 555 and 670 nm channels exceeding 50% with a standard deviation below 2.5%.

Where “Dynamic” comes from

\(T_{\text{dyn}}\) is a temperature floor that moves with the scene, computed from the scene’s own \(T_{\max}(31)\). The stated reason for it is that “the dynamic temperature scaling (Equation 4) was introduced to reduce seasonal and diurnal effects giving rise to false detection over cold land.” That is a background estimate that adapts to the scene, the same property that names the “Dynamic” in DEBRA; the sources do not state that connection themselves. Between 2004 and 2013 the background changes source rather than principle: it stops being a statistic of the current scene and becomes an external prior, a surface emissivity database plus a skin-temperature reanalysis, which is what this package implements in background.py.

Where the constants come from

On the normalization bounds, the specification says they “were determined experimentally based on a wide variety of dust case studies, with values selected toward optimizing dust contrast while maintaining an enhancement appearance consistent with the over-water algorithm.”

Two things follow, both stated rather than inferred. The bounds are empirical, fitted to case studies. And one of the fitting objectives was cross-algorithm visual continuity: the method is “composed of two algorithms (over-land and water) tuned to maintain a similar enhancement of dust crossing coastlines”, so that a dust front “maintains a similar appearance across the land/sea algorithmic boundary.” The specification also records that “performance testing for the new method is ongoing, with only minor corrections to scaling bounds anticipated”: the document describes work in progress.

Two transcription caveats

  • The specification contains an apparent slip. Discussing the weakness of Eq. 5 it reads “the Split-Window \(L_2\) terms decreases in strength for very thick dust” [sic], but per its own Table 2 the split window is \(L_1\); \(L_2\) is the 11 µm brightness temperature. The table above follows Table 2.

  • The specification refers twice to “the computer program code listing in the Appendix”, but the granted text of US 7,379,592 B2 is 18 columns of specification with 2 claims and 14 sheets of drawings, and contains no code listing. The patent as published contains no reference implementation. Whether the appendix was filed can only be settled from the image file wrapper of 10/885,526 or of the parent 10/713,908; that has not been checked.

Sources for this section

  • US 7,242,803 — “System and method for significant dust detection and enhancement”

  • US 7,379,592 — the continuation-in-part, source of every equation, table, date and quotation above

  • Miller, S. D. (2003), A consolidated technique for enhancing desert dust storms with MODIS, Geophys. Res. Lett., 30(20), 2071, doi:10.1029/2003GL018279 — cited on the US 7,379,592 face as the forthcoming GRL paper for the MODIS land/ocean enhancement

  • Miller et al. (2017), doi:10.1002/2017JD027365, as amended by the erratum of 26 February 2020 — the baseline this package implements, see Deviations

Unverified

A correction to the 2003 GRL paper is reported to have been published 2020-06-10. Neither that correction nor its relationship to the 2020 erratum on Miller et al. (2017) has been checked here, and Deviations tracks only the latter. Open item.

Verifying the current status