A Coarse-to-Fine Method for Near-Duplicate Image Retrieval with Matching Probability Model

Abstract

Image retrieval has attracted increasing interests in recent years. This paper proposes a coarse-to-fine method for fast indexing with marching probability model. We first use a vector quantized Deep Convolutional Neural Network(DCNN) feature descriptors and exploit enhanced Locality-sensitive hashing(LSH) techniques for fast coarse-grained retrieval. Then, we focus on obtaining high-precision preserved matches for fine-grained retrieval. This is formulated as a maximum likelihood estimation of a Bayesian model with latent variables indicating whether matches in the putative set are inliers or outliers. We impose the non-parametric global geometrical constraints on the correspondence using Tikhonov regularizers in a reproducing kernel Hilbert space. The problem is solved by using the Expectation Maximization algorithm. Extensive experiments on real near-duplicate images for both feature matching and image retrieval demonstrate accurate results of the proposed method which outperforms current state-of-the-art methods.