Invalidity dossier

US 6496184

Method for inferring scenes from test images and training data using probability propagation in a markov network

Current assignee: Mitsubishi Electric Research Laboratories Inc

Added 9/23/2026, 11:40:31 PM

At a glanceNo PTAB challengesNo litigation on file

Active provider: DeepSeek · deepseek-flash

Patent summary

Title, assignee, inventors, filing/issue dates, abstract, and a plain-language overview of the claims.

✓ Generated

Patent Overview: US 6,496,184 B1

Title: Method for inferring scenes from test images and training data using probability propagation in a markov network

Patent number: US 6,496,184 B1
Application number: US09/335,943
Inventors: William T. Freeman; Egon C. Pasztor; Baback Moghaddam
Original assignee: Mitsubishi Electric Research Laboratories Inc
Priority date: 1998-11-30
Filing date: 1999-06-18
Publication/grant date: 2002-12-17
Status listed: Expired – Fee Related; anticipated expiration 2018-11-30

Purpose. The invention addresses the general problem of inferring overall characteristics of a real-world signal from isolated or degraded samples—for example, estimating object velocities in video, recovering 3-D structure from 2-D images, or recovering high-resolution details from low-resolution signals. The described method aims to infer a scene from a test image.

Plain-language scope. During a training phase, labeled images and corresponding scenes are acquired (by synthesis or measurement) and partitioned into image patches and scene patches. Each patch is represented as a low-dimensional vector (the excerpt gives examples of five-dimensional scene vectors and seven-dimensional image vectors), with each vector modeled as a node in a network, preferably a Markov network; edges model statistical dependencies between nodes. In the inference phase, an unknown test image is partitioned into test image patches represented as test image vectors. Candidate scene vectors corresponding to those test image vectors are located in the training data, compatibility matrices for candidate scene vectors at neighboring network nodes are determined, and probabilities are propagated in the network until a predetermined termination condition (convergence or a fixed number of iterations) is reached, thereby inferring the scene.

What the excerpt does not establish. It does not establish litigation, ownership transfers beyond the listed assignments, validity, or prior art beyond the brief mentions of quad-tree optical flow and brightness constancy. It also omits full claim language, so the exact legal scope cannot be confirmed from this excerpt.

Source: https://patents.google.com/patent/US6496184/en

Generated 10/1/2026, 9:32:34 PM

Cases on file (0)

Specific litigation cases in our database that name US patent 6496184. The free-form analysis below may also discuss cases beyond this list.

No cases on file mention this patent. Upload a CSV or add a case manually in Admin → Manage litigation cases.

Litigation summary

Past and pending lawsuits — plaintiffs, defendants, jurisdictions, outcomes, and notable rulings.

Log in to generate
Not generated yet. Log in to request this analysis.

Proceedings on file (0)

All PTAB activity →

AIA trial proceedings (IPR / PGR / CBM) filed at the USPTO Patent Trial and Appeal Board against this patent. Sourced from the USPTO Open Data Portal and refreshed every six hours; each proceeding number deep-links to the PTAB E2E docket.

No PTAB proceedings on file. This patent has not been challenged via IPR, PGR, or CBM. The absence is itself a signal — well-asserted patents eventually attract IPRs. The LLM analysis below may surface filings the ODP feed hasn’t indexed yet.

PTAB challenges

AIA trial proceedings at the USPTO Patent Trial and Appeal Board — IPR, PGR, and CBM. Petitioners, judge panels, claim-level invalidation outcomes from Final Written Decisions, and Federal Circuit appeals. The single most important defensive datapoint after litigation history.

Log in to generate
Not generated yet. Log in to request this analysis.

Assignment history

Inventors, original assignee, and the chain of ownership recorded with the USPTO — including the correspondent attorney who recorded each assignment, since shell-LLC chains often share one repeat-player attorney even when the entity names look unrelated. Surfaces NPE / patent-troll patterns: shell-entity transfers, known asserters in the chain, repeat correspondent fingerprints, pre-litigation assignments, and bankruptcy fire-sales.

Log in to generate
Not generated yet. Log in to request this analysis.

Prior art

Earlier patents, publications, and products that may anticipate or render the claims unpatentable.

Log in to generate
Not generated yet. Log in to request this analysis.

Obviousness

Combinations of prior art that suggest the claimed invention would have been obvious under 35 U.S.C. § 103.

Log in to generate
Not generated yet. Log in to request this analysis.

Extensions

Patent term adjustments, term extensions, continuations, divisionals, family members, and expiration dates.

Log in to generate
Not generated yet. Log in to request this analysis.

Derivative works

Defensive disclosure: derivative variations of each claim designed to render future incremental improvements obvious or non-novel.

Log in to generate
Not generated yet. Log in to request this analysis.