gabor: Two Dimensional Gabor Filtering in Frequency Domain

In wvtool: Image Tools for Automated Wood Identification

Description

The function provides two dimensional Gabor function proposed by Daugman to model the spatial summation properties in the visual cortex. It returns Gabor filter in real and reciprocal space, and filtered image.

Usage

gabor.filter(x, lamda=5, theta=45, bw=1.5, phi=0, asp=1, disp=FALSE)

Arguments

x	A raster or matrix to be filtered
lamda	Wavelength of the cosine part of Gabor filter kernel in pixel. Real number greater than 2 can be used. However, lamda=2 should not be used with phase offset (phi) = -90 or 90.
theta	The orientation of parallel stips of Gabor function in degree
bw	Half responce spatial frequency bandwidth of a Gabor filter. This relates to the ratio sigma/lamda, where sigma is the standard deviation of Gaussian factor of Gabor function.
phi	Phase offset of the cosine part of Gabor filter kernel in degree
asp	Elliptcity of the Gabor function. asp=1 means circular. For asp<1 it gives elongated parallel strip
disp	If this operator is TRUE, original image, gabor filter in real space domain, that in frequency domain, and filtered image will be generated

Value

The function provides following four outputs.

kernel	151x151 gabor filter kernel
mask	A mask with kernel in the center in a space domain
freq_mask	Real part of Fourier transform of the mask in spatial frequency domain
filtered_image	Inversed Fourier transform of FFT(img)*FFT(mask)

References

J. G. Daugman, Uncertainty relation for resolution in space, spatial frequency, and orientation optimized by two d-dimensional visual cortical filters, J. Opt. Sppc. Am A, 2(7), 1160-1169, 1985

Examples

data(cryptomeria)
img <- rgb2gray(cryptomeria)
img <- crop(img,300,300)
# simple example
test <- gabor.filter(x=img, lamda=8, theta=60, bw=1.5, phi=0, asp=0.3, disp=TRUE)
## Not run: 
# azimuthal intensity distribution with respect to the orientation of Gabor function
par(mfrow=c(2,1))
Integ <- array()
filt.img <- matrix(0, nrow(img), ncol(img))
for ( i in 1:60) {
out <- gabor.filter(x=img, lamda=8, theta=3*i, bw=1.5, phi=0, asp=0.3)
filt.img <- out$filtered_img + filt.img
Integ[i] <- sum(out$filtered_img*out$filtered_img)
}
image(rot90c(filt.img),col=gray(c(0:255)/255),asp=1,axes=FALSE, useRaster=TRUE)
x <- 1:60
plot(3*x,Integ, ty="l", ylab="integrated intensity (a.u.)", xlab="azimuthal angle (deg)")

## End(Not run)

Example output

wvtool documentation built on May 1, 2019, 10:27 p.m.