R/blurpoints.R

In bbuchsbaum/eyesim: analysis of eye-movement data

#' #' blurpoints
#' #'
#' #' @param coords
#' #' @param sigma
#' #' @param xbounds
#' #' @param ybounds
#' #' @param resolution
#' #' @param weights
#' #' @param normalize
#' #' @importFrom Matrix sparseMatrix
#' #' @importFrom rflann RadiusSearch
#' #' @export
#' blurpoints <- function(coords, sigma=50, xbounds, ybounds, resolution=1, weights=rep(1, nrow(coords)), normalize=TRUE) {
#'
#'   keep <- coords[,1] >= xbounds[1] & coords[,1] <= xbounds[2] & coords[,2] >= ybounds[1] & coords[,2] <=ybounds[2]
#'
#'   coords <- coords[keep,]
#'   weights <- weights[keep]
#'
#'   xc <- seq(xbounds[1]+resolution/2, xbounds[2] - resolution/2, by=resolution)
#'   yc <- seq(ybounds[1]+resolution/2, ybounds[2] - resolution/2, by=resolution)
#'
#'   grid <- expand.grid(x=xc,
#'                       y=yc)
#'
#'   xind <- seq_along(xc)
#'   yind <- seq_along(yc)
#'
#'   igrid <- expand.grid(x=xind,
#'                        y=yind)
#'
#'   ret <- lapply(1:nrow(coords), function(i) {
#'     ret <- rflann::RadiusSearch(coords[i,,drop=FALSE], grid, (sigma*3)^2, max_neighbour=100000)
#'     ind <- ret$indices[[1]]
#'
#'     d <- sqrt(ret$distances[[1]])
#'     vals <- dnorm(d, sd=sigma) * weights[i]
#'     out <- sparseMatrix(x=vals, i=igrid[ind,1], j=igrid[ind,2], dims=c(length(xind), length(yind)))
#'   })
#'
#'   out <- Reduce("+", ret)
#'
#'   if (normalize) {
#'     out <- out/sum(out)
#'   }
#'
#'   list(im=out, x=xc, y=yc)
#' }