Gamma LAT: Reference Manual

texture

NAME
texture - Estimation of image texture defined as stdev/mean or log

SYNOPSIS
texture <data_in> <format_flag> <texture> <width> [type] [bx] [by] [weights_flag] [r_looks] [az_looks] [weights_flag] [data_in_mean]

<data_in>	(input) data file (float)
<format_flag>	format of data_in: 0: float, 1: fcomplex, 2: scomplex)
<texture>	(output) image texture (float) defined as normalized second moment
<width>	number of samples/row
[type]	texture measure flag (default=0) 0: texture defined as stdev/mean 1: log (average(i)) - average (log(i))
[bx]	estimation window size in input image range pixels (default=15)
[by]	estimation window size in input image azimuth pixels (default=bx)
[r_looks]	range multilooking (in range pixels, default=1)
[az_looks]	azimuth multilooking (in azimuth pixels, default=1)
[weights_flag]	flag for weighting function selection (default=0: constant, 1: linearly decreasing, 2: gaussian)
[data_in_mean]	(input) first moment of input data file (float) (used for an alternative texture estimation method)

EXAMPLE
texture 8059.slc_esa 2 texture2 0 750 5 25 2 10 0

Calculates the image texture (normalized second moment) of a complex valued input SAR image (in scomplex format = pairs of short integer). The estimation window size is 3 range x 15 azimuth looks. The texture image is written out using a 2 range x 10 azimuth looks multilooking.

texture 8059.slc_esa 2 texture2 1 750 5 25 2 10 0

Calculates the image texture (defined as log) of a complex valued input SAR image (in scomplex format = pairs of short integer). The estimation window size is 3 range x 15 azimuth looks. The texture image is written out using a 2 range x 10 azimuth looks multilooking

texture 8059.mli 0 texture0 750 0 5 5 1 1 0

Calculates the image texture (normalized second moment) based on the input real valued multilook SAR image. The estimation window size is 5 range x 5 azimuth samples of the multilook input image. The texture image is written out at the same sample spacing of the input image.

texture 8059.mli 0 texture0 750 0 5 5 1 1 0 8059.mli_frost

DESCRIPTION
texture estimates the texture of the input image. The image texture defined as the normalized second moment. Two texture definitions are included with the program. The first image texture type is calculated as the ratio between the standard deviation and the mean value of the values of the estimation window:

type=0: texture = stdev / mean

type=1: texture = log(average(I)) - average(log(i))

The same user defined estimation window sizes and weighting functions are used for the estimation of the mean value and the standard deviation.

If local mean values are provided in a seperate input file then the first moment is not calculated but read from this input file. This allows to previously determine the mean value using more advanced techniques, for example the MMSE technique with an adaptive estimator window size, or multi-temporal filtering. In the second approach the local mean value does not need to be constant over the estimator window used in the texture estimation. This has important advantages for the texture estimation. The estimator locally normalizes the input data values x_in using the local mean x_mean values:

x_norm = (x_in/x_mean) - 1

followed by the calculation of the second moment for the x_norm values.

Different weighting functions can be selected. For weights_flag=0 a constant weighting function is applied, i.e. all filter coefficients are set to 1. For weights_flag=1 a linear weighting function is applied, i.e. the filter coefficients decrease linearly with increasing distance to the center of the filter window. For weights_flag=2 a Gaussian weighting function is applied, i.e. the filter coefficients decrease with inclreasing distance to the center of the filter window following a Gauss function.

OPTIONS
Userdefined filter window size and weighting function.

SEE ALSO
typedef_ISP.h, average_filter, frost.