2019/04/12 by De Bortoli Valentin, Agnès Desolneux, Valentin, De Bortoli +6
Biochemistry, Genetics and Molecular Biology · Computer Science · #Cell Image Analysis Techniques #Computer Vision and Pattern Recognition (cs.CV) #FOS: Computer and information sciences #Image Retrieval and Classification Techniques #Medical Image Segmentation Techniques
paper · pdf · doi:10.48550/arxiv.1904.06428
openalex publication_date 2019/04/12 · openalex created_date 2022/03/02 · openalex updated_date 2026/07/28
In this work we introduce a statistical framework in order to analyze the\nspatial redundancy in natural images. This notion of spatial redundancy must be\ndefined locally and thus we give some examples of functions (auto-similarity\nand template similarity) which, given one or two images, computes a similarity\nmeasurement between patches. Two patches are said to be similar if the\nsimilarity measurement is small enough. To derive a criterion for taking a\ndecision on the similarity between two patches we present an a contrario model.\nNamely, two patches are said to be similar if the associated similarity\nmeasurement is unlikely to happen in a background model. Choosing Gaussian\nrandom fields as background models we derive non-asymptotic expressions for the\nprobability distribution function of similarity measurements. We introduce a\nfast algorithm in order to assess redundancy in natural images and present\napplications in denoising, periodicity analysis and texture ranking.\n