d2spec: Power Spectrum Probability Vector Calculator for Matrices
In stheoreme: Klimontovich's S-Theorem Algorithm Implementation and Data Preparation Tools

Description Usage Arguments Details Value Author(s) References See Also Examples

Function d2spec is applied to a pair of matrices (2d arrays) and generates the pair of corresponding 1d spectral power density functions (to be used then as probability vectors) specified by the method argument

1	d2spec(d2arr0, d2arr1, band=c(0,0), brks=0, method='rad', meansub=TRUE)

`d2arr0`	sample matrix
`d2arr1`	sample matrix
`band`	two border values to set a range of considered frequencies. The default c(0,0) sets full entire range from 0 to 0.5 Hz (where 1 Hz = [1/sampling_interval]) for radial method and from 0 to 180 degrees for both angular methods
`brks`	value used in a same manner as the number of cells for the histogram. The default brks=0 sets the number of cells equal to `min(length(sample0), length(sample1))/2` for radial method and equal to 180 for both angular methods
`method`	specifies the method of using the power spectrum as a probability vector. The default value 'rad' defines the calculation of spectral power density for spatial frequency space `{u,v}` as a function of spectral radius `sqrt(u^2+v^2)`. `method='ang'` to calculate spectral power density as a function of spectral angle `atan(u/v)` `method='ang90'` to calculate spectral power density as a function of shifted spectral angle `atan(v/u)`
`meansub`	logical item defining if individual baselines (as mean value) are subtracted from original matrices before application of 2d Fourier transform

Spectral power density is often being used as probability vector/matrix characterizing a state of a system, e.g. in image analysis. Here 2d fast Fourier transform algorithm is utilized to calculate power spectrum matrix P(u,v) and then the 1d spectral power density:

f(k_{i})=∑_{k_{min}}^{k_{max}}{ P(k: k_{i}<k<k_{i+1})}

where k is a metrics of frequency space defined with one of the following options: k=√{u^{2}+v^{2}} (as radius where method='rad'); k = atan(u/v) (as angle where method='ang') or k = atan(v/u) (as angle where method='ang90')

and where {k_{min},k_{max}} are defined by band argument

In general the function works similarly to d1nat. As a bonus it prints basic statistics summary for spectral power density functions alongside with technical plot. It is recommended for use as a data preparation step before following Klimontovich's S-theorem based analysis.

`f0`	spectral power density function as a probability vector representing state0 of a system
`f1`	spectral power density function as a probability vector representing state1 of a system
`efs`	vector of corresponding metrics/bin values (radial frequency or spectral angle)

Vitaly Efremov <vitaly.efremov@dcu.ie>

N.F.Zhang, A.E.Vladar, M.T.Postek, R.D.Larrabee.Spectral density-based statistical measures for image sharpness. Metrologia, 42(2005), 351-359

E.J.Groth. Probability distributions related to power spectra. Astrophysical Journal, 1975. Suppl. Ser., Vol.29, No.286, p. 285-302.

crit.stheorem, cxds.stheorem, d1spec, d1nat

# compare (harmonic+background) with (harmonic noisy period) matrices
s0<-array(2+sin(c(1:256)/3), c(16,16))
s1<-array(sin(c(1:512)+runif(512,0,2)), c(16,32))

# as radial:
b<-d2spec(d2arr0=s0, d2arr1=s1); b
b<-d2spec(s0, s1, brks=29, band=c(0.15,0.5)); b
# as angular:
b<-d2spec(s0, s1, method='ang', meansub=FALSE); b

#example of 3-step data analysis with Klimontovich's S-theorem
# Study two gratings: random vs regular
s0<-array(c(rep(0,640),rep(1,640)), c(320,320))
s1<-array(runif(5120,0,1), c(64,80))
# step a. Binarize (to make s1 comparable with s0 by its nature as a grating)
a<-utild2bin(s0, s1, method='med')
# step b. Create probability vectors as for angular space (anisotropy study)
# There is no doubt s0 is more regular
b<-d2spec(a$bin0, a$bin1, brks=36, method='ang90'); b
# step c. Compare gratings with Klimontovich's S-theorem. Renormalized entropy shift
# is negligible compared to Shannon's. Evolution from state0 to state1 is possible
# but clearly with external entropy (or energy) inflow
crit.stheorem(b$f0,b$f1)
cxds.stheorem(b$f0,b$f1)

Caution! Array1 isnt square. Will be cut!
Call:
d2spec.default(d2arr0 = s0, d2arr1 = s1)
Two spectral probability mass functions ($f0,$f1) have been generated at the common scale of radial freqs ($efs)
Statistics summary:
      expctd         var fsum xmin xmax n      mod1      mod2      mod3  H_val
f0 0.2176109 0.003670835    1    0  0.5 7 0.1785714 0.2500000 0.3214286 1.0160
f1 0.3423788 0.008044957    1    0  0.5 7 0.3928571 0.3214286 0.4642857 1.5719
     H_max
f0 1.94591
f1 1.94591
Caution! Array1 isnt square. Will be cut!
Call:
d2spec.default(d2arr0 = s0, d2arr1 = s1, band = c(0.15, 0.5), 
    brks = 29)
Two spectral probability mass functions ($f0,$f1) have been generated at the common scale of radial freqs ($efs)
Statistics summary:
      expctd         var fsum xmin xmax  n      mod1      mod2      mod3  H_val
f0 0.2449084 0.003234614    1 0.15  0.5 29 0.2163793 0.2887931 0.3612069 0.7968
f1 0.3281579 0.005721408    1 0.15  0.5 29 0.3612069 0.2887931 0.4336207 1.3674
      H_max
f0 3.367296
f1 3.367296
Caution! Array1 isnt square. Will be cut!
Call:
d2spec.default(d2arr0 = s0, d2arr1 = s1, method = "ang", meansub = FALSE)
Two spectral probability mass functions ($f0,$f1) have been generated at the common scale of spectral angles ($efs)
Statistics summary:
      expctd       var fsum xmin xmax   n  mod1  mod2  mod3  H_val    H_max
f0  93.68743  249.7329    1    0  180 180  90.5 153.5 161.5 0.3164 5.192957
f1 128.08069 2846.0915    1    0  180 180 156.5 164.5 171.5 2.7694 5.192957
Caution! Array1 isnt square. Will be cut!
Call:
d2spec.default(d2arr0 = a$bin0, d2arr1 = a$bin1, brks = 36, method = "ang90")
Two spectral probability mass functions ($f0,$f1) have been generated at the common scale of spectral angles ($efs)
Statistics summary:
     expctd      var fsum xmin xmax  n mod1  mod2  mod3  H_val    H_max
f0 92.50000    0.000    1    0  180 36 92.5  82.5  97.5 0.0000 3.583519
f1 90.65819 2523.968    1    0  180 36 47.5 137.5 132.5 3.5446 3.583519

 S-theorem convergence criterion

System evolution from state0 to state1 is thermodynamically possible through an indirect
 medium state2 (R^2 = 0.9604).


 S-theorem open system evolution model

Overall entropy shift {H1-H0 = dS + dI}:
3.54462331081128 {0to1} = 0.0639352379338206 {0to1} + 3.48068807287746 {0to1}
where dS consists of two:
0.0639352379338206 {0to1} = 0.0639352379338196 {0to2} + 1e-15 {2to1}