Method and apparatus for automatic image categorization using image texture
AssigneeSeiko Epson Corporation
InventorHui Zhou
Life of this patent
Estimating the term…
Abstract
A method of automatically categorizing an input image comprises extracting texture features of the input image and generating a signature vector based on extracted texture features. The generated signature vector is processed using at least one classifier to classify the input image.
Claims15
A method of automatically categorizing an input image, comprising:
- using a processing unit to:
- extract texture features of said input image based on localized edge orientation coherence and using a non-parametric Markov Random Field (MRF) model and generating a signature vector based on said extracted texture features;
- and
- process said signature vector using at least one classifier to classify said input image.
A categorization system for automatically categorizing an input image comprising:
- a signature vector generator operative to extract texture features of said input image and generate a signature vector based on extracted features;
- and
- a processing node network operatively coupled to said signature vector generator and adapted to process said generated signature vector using at least one classifier to classify said input image;
- and wherein
- said signature vector comprises:
- an edge orientation component;
- and
- a texture component;
- and
- said signature vector generator generates the edge orientation component based on local edge orientation coherence of pixels in said input image and generates said texture component based on a non-parametric Markov Random Field (MRF) texture model.
A method of automatically categorizing an input image, comprising:
- using a processing unit to:
- pre-process said input image to form a gray-scale image;
- generate a first edge image by performing edge detection on said gray-scale image;
- calculate an intensity threshold value from said first edge image;
- apply said threshold value to said first edge image to generate a thresholded edge image;
- and
- process said thresholded edge image using a non-parametric Markov Random Field (MRF) model to generate texture features for categorizing said input image.
A non-transitory computer-readable medium embodying machine-readable code for categorizing an input image, said machine-readable code comprising:
- machine-readable code for extracting texture features of said input image based on localized edge orientation coherence and using a non-parametric Markov Random Field (MRF) model and generating a signature vector based on said extracted texture features;
- and
- machine-readable code for processing said signature vector using at least one classifier to classify said input image.
Dependencies are inferred from the claim text.