About digital number, linear stretching and vegetation indices

[emani]

Does the value of digital number change after linear stretching? And is it fine to obtain vegetation indices from linerly stretched images? 

 

Thanks

Emani

Edited May 12, 2014 by emani

[30000]

Histogram stretching does not change the DNs of an image, merely how the image is displayed. Vegetation indices will be calculated from the DN values of the bands themselves, not how it is displayed, so it is fine to use NDVI or any other vegetation index while using linear stretching. You can confirm this for yourself by inquiring a certain pixel, then changing the histogram to various stretches. The file pixel values for each band will stay the same, while the look up table values will changed based on the stretching.

[emani]

Histogram stretching does not change the DNs of an image, merely how the image is displayed. Vegetation indices will be calculated from the DN values of the bands themselves, not how it is displayed, so it is fine to use NDVI or any other vegetation index while using linear stretching. You can confirm this for yourself by inquiring a certain pixel, then changing the histogram to various stretches. The file pixel values for each band will stay the same, while the look up table values will changed based on the stretching.

Hi, Thanks for the reply....I actually stretched the image in R using this code which was given to me by my friend and now I see that my DN values have changed as I checked them on ERDAS. As digital number should be an Integer but all the values I got after running this code were fractions. 

I am posting the code which I used below.

rm(list=ls(all=TRUE))
require(rgdal)
require(rgl)
require(maptools)

StartPath <- getwd()

#Defining the root directory
Root <- "E:/Programming/Emani"

Path_in <- paste(Root, "/Input",sep="")
Path_out <- paste(Root, "/Output",sep="")

dir.create(Path_out,showWarnings=FALSE, recursive=TRUE)

hist.strecth <- function(x)
{
	n <- dim(x)[2]
	for(k in 1:n)
	{
		data <- x[,k]
		cur.lim <- quantile(data,c(0.025,0.975),na.rm=TRUE)
		data <- pmax(cur.lim[1],pmin(cur.lim[2],data))
		data <- floor(255*(data-cur.lim[1])/(cur.lim[2]-cur.lim[1]))
		data[is.na(data)]<-0
		x[,k] <- data
	}
	return(x)
}
#======================================================================================
# Read image
#======================================================================================
image.fn <- "LAUW_20130327"

A <- readGDAL(paste(Path_in,"/",image.fn,".tif",sep="")) 

Nb <- dim(A@data)[2]

#summary(A@data)

# Read vector file, for subsetting

vector.fn <- "extent"

V <- readOGR(paste(Path_in,"/",vector.fn, sep=""), layer=vector.fn)

i<-1
j<-1

polxy <- coordinates(V@polygons[[i]]@Polygons[[j]])

xy <- coordinates(A)

tmp <- point.in.polygon(xy[,1],xy[,2],polxy[,1],polxy[,2])
ind <- which(tmp==1)
indout<- which(tmp!=1)




minDN <- array(0,Nb)
maxDN <- array(0,Nb)

for(k in 1:Nb)
{
	x <- A@data[ind,]
	minDN[k] <- min(x[,k])
	data <- x[,k]
	maxDN[k] <- quantile(data,0.99,na.rm=TRUE)
}

Acalibrated <- A

for(k in 1:Nb)
{
	x <- A@data[ind,k]
	x <- x-minDN[k]
	x <- x*255/(maxDN[k]-minDN[k])
	
	x[x<0] <- 0
	x[x>255]<-255
	
	Acalibrated@data[ind,k] <- x
}


Acalibrated@data[indout,] <- 0
hist(Acalibrated@data$band4)

# output MRF solution to a .tif file
OUT <- Acalibrated
setwd(Path_out)

imagefn.out <- paste(image.fn,"_calibrated.tif",sep="")

OUT.tif<-create2GDAL(OUT,drivername="GTiff",type="Float32",mvFlag=-1)

saveDataset(OUT.tif,imagefn.out)
GDAL.close(OUT.tif)




setwd(StartPath)

# The End

thanks

emani