LACD

Local adaptive change detection


EnvironmentsPYTHON :: EASI :: MODELER
Quick linksDescription :: Parameters :: Parameter descriptions :: Details :: Example :: Algorithm :: References :: Related

Back to top

Description


LACD calculates a difference image from two input images.
Back to top

Parameters


Name Type Caption Length Value range
FILI_IN* String Input file name (Iin = I1) 1 - 192  
FILI_REF* String Input file name (Iref = I2) 1 - 192  
FILO_DIF* String Output LACD image file name 1 - 192  
DBIC_REF* Integer Input reference channels 1 - 1  
DBIC_IN* Integer Input raster image channel 1 - 1  
DBOC_DIF* Integer Output difference channel 1 - 1  
KSIZE Integer Kernel size 0 - 1 1 -
Default: 7
MONITOR String Monitor mode 0 - 3 ON, OFF
Default: ON

* Required parameter
Back to top

Parameter descriptions

FILI_IN

Specifies the PCIDSK file that contains the input image data.

FILI_REF

Specifies the PCIDSK file that contains the reference image data.

FILO_DIF

Specifies the name of the output PCIDSK file to receive the LACD difference image. This file must already exist in the working directory.

DBIC_REF

Specifies the input reference image channel from the reference image file.

The reference image channel must be the same size and datatype (8-bit unsigned) as the input raster channel (DBIC_IN).

DBIC_IN

Specifies the input raster image channel from the input file.

This image channel must be the same size and datatype (8-bit unsigned) as the reference image channel (DBIC_REF).

DBOC_DIF

Specifies the output image channel to receive the Local Adaptive Change Detection data.

The output difference channel must be the same size as the input channel (DBIC_IN) and the input reference channel (DBIC_REF). The output channel datatype must be 32-bit real (32R).

KSIZE

Specifies the size of the square kernel M x M over which the local coefficients of b0 and b1 are determined, where M=(2*KSIZE+1).

The size of M defines the upper boundary of the size of the change that will be detected. The length of time to compute is linearly proportional to M**2.

MONITOR

The program progress can be monitored by printing the percentage of processing completed. A system parameter, MONITOR, controls this activity.

Available options are:

Back to top

Details

LACD performs a change detection comparison of two input images, and writes the output to a 32-bit real image of the same size. Input is restricted to 8-bit unsigned images, of the same X/Y (pixels and lines) dimensions.

LACD performs its change detection on a localized scale. The least-squares minimization parameters, B0 and B1, are calculated using local windows of data specified by the Kernel Size (KSIZE). The formula for the difference calculation is the same as for GCD, except that the B0 and B1 parameters change with each pixel:

FILI_DIF = Minimize || FILI_IN - (B1 * FILI_REF + B0) ||
Back to top

Example

Compute the local adaptive change between the 2nd channel of the multispectral image database (ms.pix), with the 1st channel of the panchromatic image database (pan.pix). Place the output image into the 1st channel of LACD.pix. Set the kernel size to 7 so that the sliding window is 15 X 15.

EASI>FILI_IN  = "ms.pix"
EASI>FILI_REF = "pan.pix"
EASI>FILO_DIF = "lacd.pix"
EASI>DBIC_IN  = 2
EASI>DBIC_REF = 1
EASI>DBOC_DIF = 1
EASI>KSIZE    = 7

EASI> run LACD
Back to top

Algorithm

Local Adaptive Change Detection obtains a difference image from two input images, I1 and I2, by the the following formula:

I1 -  (b1 * I2 + b0)            (1)

The coefficients, b0 and b1, were calculated by using Least Square Minimization; for example:

Min || I1 -  (b1 * I2 + b0) ||  (2)

The coefficients b0 and b1 are calculated locally from a sequence of sliding M x M windows, pixel by pixel, where M = 2*KSIZE + 1. Equation (1) is then applied to obtain the difference image of M x M locally.

Back to top

References

Jack S. Li, (Revised by Victor Tom and Stephen Bento), 1996. "Local Adaptive Change Detection, Atlantic Aerospace Electronic Corporation", Waltham, MA , USA.

© PCI Geomatics Enterprises, Inc.®, 2024. All rights reserved.