Codes‎ > ‎

Cross-sensor relative spectral alignment

This page presents data and codes to reproduce the results in: 

Volpi, Michele; Camps-Valls, Gustau; Tuia, Devis (2015); Spectral alignment of multi-temporal cross-sensor images with automated kernel canonical correlation analysis; ISPRS Journal of Photogrammetry and Remote Sensing, vol. 107pp. 50-63, 2015


The remote sensing community is particularly interested in general and automatic models able to integrate cross-sensor acquisitions for land cover classification and in particular for change detection purposes. However, freely available datasets and tools / codes are truly missing to foster the development of such techniques. In this page, we distribute the data and MATLAB codes used to produce the results presented in [1].

On September 4, 2011, the region has been struck by “the most destructive wildland-urban interface wildfire in Texas history”, which counts 2 casualties, more than 1300 destroyed buildings and almost burned entirely the Bastrop county state park. For more details, read the report on the Bastrop county website [2] or see the related Wikipedia pageThis dataset is composed of a set of four images acquired by different sensors over the Bastrop County, Texas (USA). It is composed by a Landsat 5 TM as the pre-event image and a Landsat 5 TM, a EO-1 ALI and a Landsat 8 as post-event images. We provide also the ground truth we prepared to evaluate the method.

The MATLAB code implements the fully automatic cross-sensor relative spectral alignment technique presented in the paper. It relies on the automatic model selection for the kernel canonical correlation analysis transformation to find directions of cross-correlation between the data. Change detection is then evaluated using the change vector analysis. For more details, please see the paper. We also make available the cross-sensor (which includes single sensor bi-temporal images) on which we evaluated the proposed system.


Sensor 1

Post-event Sensor 2

Obtained change detection


Pre-event transformed
Post-event transformed
Difference image transformed 


These data are distributed by the Land Processes Distributed Active Archive Center (LP DAAC), located at USGS/EROS, Sioux Falls, SD. Further details may be found here.

IF YOU ARE USING THIS DATA PLEASE ACKNOWLEDGE NASA'S LP DAAC TO SUPPORT THEIR EFFORTS! If you are using the code and/or the ground truth we are providing along the data, please make sure to reference [1] as well. Thanks! 

Help research, redistribute data and codes and receive credit for your work!

Our work is shared under a creative common license! Please check it!


CROSS-SENSOR ALIGNMENT MATLAB CODE. Please note that in this RELEASE version there are codes from other external sources. If you are reusing the code we provide, please include the acknowledgments in your files! 
By running RELEASE_MAIN_CrossSensorAlignment.m you should obtain (roughly) the same numbers we have in our paper for the BASTROP case study. More details are included in the code. Be sure to set up and check RELEASE_Config.m with your local details.

CROSS-SENSOR ALIGNMENT DATASET. Put the data in the ./imm/ folder, as is. The code will do the rest. Check RELEASE_ReadData.m to see how the data is organized. Images are in uint8 and uint16 format, the ground truth is in binary logical format.

For any bug, comment or request, feel free to contact me at michele ( a dot ) volpi ( an at ) ed ( a dot ) ac ( a dot ) uk

[1] Volpi, Michele; Camps-Valls, Gustau; Tuia, Devis (2015); Spectral alignment of multi-temporal cross-sensor images with automated kernel correlation analysis; ISPRS Journal of Photogrammetry and Remote Sensing, vol. 107, pp. 50-63.
[2] Ridenour, Karen; Rissel, Sean; Powell, Wade; Gray, Rich; Fisher, Mike; Sommerfeld, Julie (2012); The Bastrop Complex Wildfire: The most destructive wildland urban interface wildfire in Texas history -- A case study; Techical Report, Bastrop Complex Wildfire Case Study, Texas Forest Service and Bastrop County Office of Emergency Management, available online at