Image Labeling Model Based on Conditional Random Fields

Abstract

We present conditional random fields (CRFs), a framework for building probabilistic models to segment and label sequence data, and use CRFs to label pixels in an image. CRFs provide a discriminative framework to incorporate spatial dependencies in an image, which is more appropriate for classification tasks as opposed to a generative framework. In this paper we apply CRF to an image classification tasks: an image labeling problem (manmade vs. natural regions in the MSRC 21-object class datasets). Parameter learning is performed using contrastive divergence (CD) algorithm to maximize an approximation to the conditional likelihood. We focus on two aspects of the classification task: feature extraction and classifiers design. We present classification results on sample images from MSRC 21-object class datasets.