In this communication, we show that two simple assumptions on the covariances matrices of color images, namely stationarity and commutativity, can explain the observed shape of decorrelated spatiochromatic elements (bases obtained by PCA) of natural color images. The validity of these assumptions is tested on a large database of RAW images. Our experiments also show that the spatiochromatic covariance decays exponentially with the spatial distance between pairs of pixels and not as a power law as it is commonly assumed.
CITATION STYLE
Provenzi, E., Delon, J., Gousseau, Y., & Mazin, B. (2015). On spatiochromatic features in natural images statistics. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 9280, pp. 46–55). Springer Verlag. https://doi.org/10.1007/978-3-319-23234-8_5
Mendeley helps you to discover research relevant for your work.