Light Field Retrieval In Compressed Domain

Abstract

In this paper, we present a novel approach for light field retrieval in the transformed domain. The light field data set is characterized with low wavelet transform coefficients. Our algorithm first applies a wavelet transform on 20 key images of the light field structure. The low wavelet transform coefficients are then quadtree decomposed, into a set of homogeneous blocks. Each set of homogeneous blocks represent an object in the considered image. We use texture and color features to characterize the object image; the similarity between two images is measured by matching their histograms. The experimental results and comparisons show the performance of the proposed technique.