Search research articles

Search research articles

Related Experiment Videos

Style consistent classification of isogenous patterns.

Prateek Sarkar¹, George Nagy

¹Palo Alto Research Center, 3333 Coyote Hill Road, Palo Alto, CA 94304, USA. psarkar@parc.com

IEEE Transactions on Pattern Analysis and Machine Intelligence

|January 5, 2005

Summary

This summary is machine-generated.

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by

Same author

Viscosity-responsive styryl benzoxazole probes for lipid droplet visualization.

Organic & biomolecular chemistry·2026

Same author

Styryl benzoxazolium salts as environment-sensitive mitochondrial probes for imaging ferroptosis.

Organic & biomolecular chemistry·2025

Same author

Comment: projection methods require black border removal.

IEEE transactions on pattern analysis and machine intelligence·2009

Same author

A comparative study of local matching approach for face recognition.

IEEE transactions on image processing : a publication of the IEEE Signal Processing Society·2007

Same author

Analytical results on style-constrained bayesian classification of pattern fields.

IEEE transactions on pattern analysis and machine intelligence·2007

Same author

Style context with second-order statistics.

IEEE transactions on pattern analysis and machine intelligence·2005

ABOUT JoVE

Overview Leadership Blog JoVE Help Center

AUTHORS

Publishing Process Editorial Board Scope & Policies Peer Review FAQ Submit

LIBRARIANS

Testimonials Subscriptions Access Resources Library Advisory Board FAQ

RESEARCH

JoVE Journal Methods Collections JoVE Encyclopedia of Experiments Archive

EDUCATION

JoVE Core JoVE Business JoVE Science Education JoVE Lab Manual Faculty Resource Center Faculty Site

Terms & Conditions of Use

Style constrained classifiers improve pattern recognition accuracy by modeling feature dependencies within pattern groups (fields) sharing a common origin. This method reduced classification errors by nearly 25 percent in experiments with printed digits.

Area of Science:

Pattern recognition
Machine learning
Computer vision

Background:

Patterns often appear in groups (fields) with a shared origin, leading to style consistency and statistical dependencies among their features.
Existing classifiers often treat patterns individually, ignoring the valuable information conveyed by group-level style consistency.

Purpose of the Study:

To develop and evaluate a style-constrained classifier that leverages statistical dependencies within pattern fields for improved accuracy.
To model the effects of style consistency on pattern features using hierarchical mixture models.

Main Methods:

Developed a hierarchical mixture model where each field originates from a mixture of styles, and patterns within a field follow a class-style conditional mixture of Gaussians.
Proposed an optimal style-constrained classifier that processes entire fields of patterns, considering their shared, unknown style.

Related Experiment Videos

Utilized concatenated pattern features (field-features) to capture style effects.

Main Results:

Style-constrained classification significantly reduced errors on fields of printed digits by nearly 25 percent compared to singlet classifiers.
The proposed method's performance advantage increases with the length of the pattern fields, as longer fields provide more information about the underlying style.

Conclusions:

Modeling style dependencies within pattern fields is crucial for enhancing classification accuracy in pattern recognition.
The hierarchical mixture model provides an effective framework for style-constrained classification, outperforming traditional singlet approaches.
The effectiveness of style-constrained classification is directly related to the amount of information available about the shared style, favoring longer pattern fields.