.png){kind=icon}
.png){kind=icon}
Return
Corrective pattern-matching simulation with controlled local-mean histogram
.png){kind=icon}
PRE
AI
DOI：10.1007/s00477-014-0864-9.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
This paper presents a new stochastic simulation method that improves on our unilateral patchwork simulation method (Faucher et al., Stoch Environ Res Risk Assess, 27:253-273, 2012) by eliminating anisotropy biases. As in our unilateral method, images are built by assembling patterns together while controlling the local-mean histogram. The patterns, which are square image-pieces, are picked in a reference image. The reference image is used as a data bank holding the statistical informations about the random field to simulate. In contrast with the unilateral method, the path followed by our new simulation is random and guided by local errors. The new method, called corrective pattern-matching simulation, proceeds iteratively by making local corrections to the simulated image. For several types of images, it is shown that our simulations respect conditioning data and reproduce faithfully the reference image visual appearance. It is shown that the control of the local-mean histogram allows to control one-point statistics and multi-point statistics.
Keywords:
Simulations with patterns
Reference image
Histogram control
Pattern similarity
Multiple-point statistics
Geostatistics
Greedy algorithm
Journal
IF:
3.6
Papers:
3.5K
Citations:
6.9K