Structural Texture Similarity Metrics for Image Analysis and Retrieval

• Published in: IEEE transactions on image processing Vol. 22; no. 7; pp. 2545 - 2558
• Main Authors: Zujovic, Jana, Pappas, T. N., Neuhoff, D. L.
• Format: Journal Article
• Language: English
• Published: New York, NY IEEE 01.07.2013; Institute of Electrical and Electronics Engineers
• Subjects: Algorithms; Applied sciences; Correlation; Databases, Factual; Exact sciences and technology; Gray-scale; Humans; Image coding; Image processing; Image Processing, Computer-Assisted - methods; Information theory; Information, signal and communications theory; Measurement; Models, Theoretical; Natural textures; Pattern recognition; perceptual quality; PSNR; Signal and communications theory; Signal processing; Signal representation. Spectral analysis; Signal, noise; statistical models; Surface Properties; Telecommunications and information theory; Visual Perception; Performance evaluation; Texture analysis; State of the art; Subband decomposition; Query; Similarity; Image retrieval; Information retrieval; Texture; Subjective evaluation; Implementation; Visual perception; Natural textures; Image analysis; Statistical model; statistical models; Database; Signal processing; Metric; perceptual quality; Pattern synthesis; Content analysis; Signal to noise ratio
• ISSN: 1057-7149; 1941-0042
• DOI: 10.1109/TIP.2013.2251645

We develop new metrics for texture similarity that accounts for human visual perception and the stochastic nature of textures. The metrics rely entirely on local image statistics and allow substantial point-by-point deviations between textures that according to human judgment are essentially identical. The proposed metrics extend the ideas of structural similarity and are guided by research in texture analysis-synthesis. They are implemented using a steerable filter decomposition and incorporate a concise set of subband statistics, computed globally or in sliding windows. We conduct systematic tests to investigate metric performance in the context of "known-item search," the retrieval of textures that are "identical" to the query texture. This eliminates the need for cumbersome subjective tests, thus enabling comparisons with human performance on a large database. Our experimental results indicate that the proposed metrics outperform peak signal-to-noise ratio (PSNR), structural similarity metric (SSIM) and its variations, as well as state-of-the-art texture classification metrics, using standard statistical measures.