Converting ENVI files to & from MATLAB format
Most of the analysis was carried out using MATLAB routines, whereas reading the ALOS data was performed using ENVI with filtering in the associated IDL. To change between ENVI and MATLAB format I used the two routines enviread.m and enviwrite.m. These required a little modification to get the correct orientation on output. For complex data use enviread_cpx.m. They were obtained in their original form from the mathworks website.
Multi-temporal analysis of ScanSAR images
The ALOS usually acquired ScanSAR images of the same scene every 46 days and in this work we used stacks of up to 12 images for the same scene covering a period of ~18 months. We showed in the first paper that best way to identify forest change was to analyse these simply by finding the temporal standard deviation for each pixel. This is not a procedure for identifying natural forest and the analysis needs to be carried out only over the known forest areas. A description of the procedure is described in the pdf document ScanSAR analysis and MATLAB codes used in the analysis can be downloaded here.
Analysis of Fine Beam Dual (FBD) images
FBD images of the same scene were obtained roughly bi- or tri-annually by the ALOS. The first paper describes the analysis of time-separated pairs FBD intensity images by a ratio technique and the second paper extends these methods to the analysis of texture. A description of the procedure is given in the pdf document FBD analysis and MATLAB codes used in the analysis are given here. Although multi-channel filtering of .mli files was usually performed using IDL the equivalent Matlab code is available here.
Combination of ScanSAR and FBD detection schemes
We found that the analysis of FBD scenes benefited from the combination of intensity and textural information. Further improvements in the detection rates could be obtained by combination with the SCanSAR results. Data Fusion methods and Principle component analysis (PCA) were tried. Data fusion is a quick and simple method to apply and in some cases gave the better results, but overall the best results were obtained using PCA. Matlab code to perform data fusion and PCA (using MATLAB routines princomp.m and zscore.m) are given with the FBD_analysis files and described in the document FBD analysis. To use these techniques ScanSAR SD results must first be co-resistered with the FBD footprint as decribed in the pdf document ScanSAR-FBD combination. For this I used Envi and Matlab.