The best way to compare changes from the images of landscapes is a still image processing in special programs . ENVI, ERDAS Imaging, IDRISI, Definiens (eCognition), etc. For example, in ENVI has a special module which detects changes by comparing satellite images at different times . Selecting Landsat 4-5 Landsat or Landsat 7 or 8 depends on the purpose and scope of research. Spatial resolution Landsat 4-5 - 30m , Landsat 7 , Landsat 8 can be adjusted up to 15 m , due to the panchromatic image. Radiation resolution Landsat 4,5,7 8 bits, and Landsat 8 - 16 bits. Additionally, Landsat 7 , 2003 after a lower quality , due to failure of the scanner and the bands lack of information.

First of all, to compare different images of Landsat 5 and Landsat 7 (and Landsat 8 too), you should correct the images doing an radiometric correction. This procedure is quite time-consuming, but it is necessary to remove the atmosphere effect, to avoid sensor errors or distorsions, and to be able to campare two images of different sensors.

First of all thank you very much for your answers. I Use ENVI 5 to convert landsat images from DN to Radiance and Reflectanse and to apply Dark Substract (The script that has the program). I try to find papers with similar subject, about comparing. I am not sure if I need to execute PCA or something else.

Vassilis,

In addition to previous contributions, I recommend you the article "Radiometric cross-calibration of the Landsat-7 ETM+ and Landsat-5 TM sensors based on tandem data sets" by Teillet et al (2001). There is also a recent paper by Kovalskyy and Roy, published in RSE, that could be of interest (see on internet). However, it doesn't deal with image processing issues, which I understand is what you are looking for.

Wht do you want to compare, simply the radiometric data or looking for change detection?

Comparison between remote sensing systems takes the following aspects:

1. Spatial resolution

2. Spectral resolution

3. Radiometric resolution

4. Temporal resolution

5. Areal coverage

but if you want to make change detection between two image dates, it is better to do this after image classification

A comment on Mr. Uddin answer: For adequate change detection evaluation, the cross-calibration among sensors is mandatory and must be performed in the preprocessing phase; otherwise, artefacts caused by their different technical characteristics may be interpreted as changes.

Dear Ruxanda thanks for the article. I want to compare images and looking process from a theoretical point of view. Also when I do the calibration of images I'm trying to anderstand if it is enough to see the difference between NDVI of both photos and same scater plots of the bands?

Dear Saif yes I would like to compare radiometric data. Change detection I will try in later time (maybe).

Vassilis,

Cross-calibration is performed on raw data (before computing NDVI) and is generally assured by satellite owners (NASA for Landsat). This is a dynamic procedure and it is possible that for certain periods the sensors are nor properly calibrated and/or cross-calibrated (see the attached paper). Generally, this is not a problem to be solved at the user level; nonetheless, many times it is detected by users, especially from the academic community. Coming at your current position: (1) read everything available to you on satellite sensor calibration ( + focus on Landsat 5 and 7); (2) discuss radiometric calibration (including cross-calibration) with knowledgeable people (e.g., NASA or your data provider) with respect to the images you have; (3) process NDVI on radiometric corrected and inter-calibrated data; (4) perform co-registration and change detection (several algorithms are available; I recommend you to try the more you can); (5) when you interpret the changes, keep in mind that some of them may be artefacts produced by data and model uncertainties; (6) try to statistically validate with independent data; if it is possible to you, treat "suspect changes" separately to detect if they are artefacts.

I also suggest you to read the paper "Evaluation of SPOT/HRV data over temporal series acquired during the ADAM project" by Oro, Baret and myself (on Researchgate). You will have a glimpse about temporal and spatial coherence of series of satellite data acquired by different sensors. Good luck !

Dear Ruxandra thank you very mach about your answer.

Addition from the article you sent me and I found the following which does something similar from what I want. A COMPARATIVE ANALYSIS OF CBERS AND LANDSAT DATA (http://www.asprs.org/a/publications/proceedings/baltimore09/0005.pdf)

Bravo, keep on trace !