You have struck on a major problem in creating historical overlays and detailed studies of geographical change.  It is easy to do in an approximate way.  It is very difficult to do accurately.  There are several issues: 1) imaging the maps; 2) georeferencing them; and 3) as a subset of two, map projection details and geomatic information.  First, you need to know how accurate your original maps are -- not just the scale.  Do they conform to some positional map accuracy standard?  Often a national map agency will have these standards, even though they do not test every map they publish to see if it conforms.  (It is expensive to validate the accuracy of a map sheet.) Then the map is scanned by a scanner and you need to know the scanner's accuracy of scanning because a pretty accurate map can be distorted in the scanning process.  Many digital collections implicitly assume the digital version of a map that conforms to a map accuracy standard also has the same accuracy.  It doesn't.  So you have to ask:  What is the error embodied in the scan?  If the maps were scanned by sheet fed scanners, what is the error of scanning in the stepper motor dimension (as the map moves through the scanner)?  Many scanners don't tell, or the manufacturer didn't measure it.  Just because it looks good to the eye does not mean it is accurate.  You need to find a flat-bed scanner with known and extremely low error in imaging.  We solved this at the W.E. Upjohn Center with custom flat-bed scanners that image a map to plus/minus 2/10,000ths of an inch accuracy.  If all your maps are imaged to the same (extremely high) accuracy standard, it will be easier to work with historical overlays.  Second, one needs to know in detail all the map projection and geomatic info by which to georeference the maps.  Do not use maps that have been "crowd georeferenced" as at a number of public map collections.  This is only a way to get "quick and dirty" solutions.   Third, using generic information is okay -- but most likely it will be an approximate solution.  This is a problem in historical maps because often they do not give anything more than a general statement about the map projection.  

    This is why we at the W.E. Upjohn Center created an "Authoritative U.S. Topographic Maps" (C) initiative to image and rectify all U.S. government issued maps back to the era of Thomas Jefferson.  To date, two series of maps have been completed and will be downloadable onto mobile devices for crowd-sourcing and updating -- ca. 114,000 maps of the U.S.  The Center received an outstanding map product award from the International Map Industry Association in 2013 for these maps.  (This Cener, incidentally, would be happy to work with agencies that desire to create digitally as-accurate-as-possible map series.)

    How to solve the problem in your case, though?  Probably you could hunt for really high quality GPS points (usually taken by LIS surveyors, etc.)  These points you could use to geo-reference key landmarks or unchanging topographic features.  Don't use just any GPS readings, though.  They have error in them too -- so read up.  And definitely, do not fit it to Google Earth imagery (plus/minus up to 15 m.).  This is good only for small scale maps -- and, of course, is a delight to the eye to see (but not to measure).  Also, rubber sheeting needs to be done mindfully, else one unwittingly introduces error of unknown dimension in the parts of the map image not proximate to the selected points.

    Good luck in your endeavor!

You can use Q GIS software , open your existing file in this software. In the main tool bar go to web option , click link with google earth you can compare your map with google earth.

You need one georefernced base map for rectifing all your historical maps and current map using spatial adjustment tool. It is available in ArcGis software. You find well defined position in all map then give point for spatial adjustment. Then overlay all maps and find out which place is not siting properly then again u give more points.

You can check this publication, the methodology is based on artificial neural network. According with your resources, you will get a transformation with reasonable accuracy.

Article Planes coordinates transformation between PSAD56 to SIRGAS u...

You have struck on a major problem in creating historical overlays and detailed studies of geographical change.  It is easy to do in an approximate way.  It is very difficult to do accurately.  There are several issues: 1) imaging the maps; 2) georeferencing them; and 3) as a subset of two, map projection details and geomatic information.  First, you need to know how accurate your original maps are -- not just the scale.  Do they conform to some positional map accuracy standard?  Often a national map agency will have these standards, even though they do not test every map they publish to see if it conforms.  (It is expensive to validate the accuracy of a map sheet.) Then the map is scanned by a scanner and you need to know the scanner's accuracy of scanning because a pretty accurate map can be distorted in the scanning process.  Many digital collections implicitly assume the digital version of a map that conforms to a map accuracy standard also has the same accuracy.  It doesn't.  So you have to ask:  What is the error embodied in the scan?  If the maps were scanned by sheet fed scanners, what is the error of scanning in the stepper motor dimension (as the map moves through the scanner)?  Many scanners don't tell, or the manufacturer didn't measure it.  Just because it looks good to the eye does not mean it is accurate.  You need to find a flat-bed scanner with known and extremely low error in imaging.  We solved this at the W.E. Upjohn Center with custom flat-bed scanners that image a map to plus/minus 2/10,000ths of an inch accuracy.  If all your maps are imaged to the same (extremely high) accuracy standard, it will be easier to work with historical overlays.  Second, one needs to know in detail all the map projection and geomatic info by which to georeference the maps.  Do not use maps that have been "crowd georeferenced" as at a number of public map collections.  This is only a way to get "quick and dirty" solutions.   Third, using generic information is okay -- but most likely it will be an approximate solution.  This is a problem in historical maps because often they do not give anything more than a general statement about the map projection.  

    This is why we at the W.E. Upjohn Center created an "Authoritative U.S. Topographic Maps" (C) initiative to image and rectify all U.S. government issued maps back to the era of Thomas Jefferson.  To date, two series of maps have been completed and will be downloadable onto mobile devices for crowd-sourcing and updating -- ca. 114,000 maps of the U.S.  The Center received an outstanding map product award from the International Map Industry Association in 2013 for these maps.  (This Cener, incidentally, would be happy to work with agencies that desire to create digitally as-accurate-as-possible map series.)

    How to solve the problem in your case, though?  Probably you could hunt for really high quality GPS points (usually taken by LIS surveyors, etc.)  These points you could use to geo-reference key landmarks or unchanging topographic features.  Don't use just any GPS readings, though.  They have error in them too -- so read up.  And definitely, do not fit it to Google Earth imagery (plus/minus up to 15 m.).  This is good only for small scale maps -- and, of course, is a delight to the eye to see (but not to measure).  Also, rubber sheeting needs to be done mindfully, else one unwittingly introduces error of unknown dimension in the parts of the map image not proximate to the selected points.

    Good luck in your endeavor!

Have a look at MapAnalyst

You need to locate control points with unchanged position in both the historical map and an updated topographic map complying to map accuracy standards. Such points could be the top of a hill or near the top of a hill, intersections, buildings, monuments, etc. You must locate as many control points as possible covering uniformly the entire area. Then use part of the control points for georeference and the rest to evaluate the accuracy. Use various degree (1-3) polynomials for the transformation of coordinates and if the RMS of the transformation results is the desirable one, then do resampling to have the historical map in the desirable coordinate system with desirable accuracy.

There are some useful and thoughtful warnings above about simple tick point co-registration of maps.  The process using polynomials is described in a paper Paul Mather and I wrote ages ago:

D.J. Unwin and P.M. Mather (1998) Selecting and using ground control points in image rectification and registration, Geographical Systems, 5(3), 239-260

The rule is to be very careful and do not trust your GIS to do it properly.

Normally, we do not know the projection of historical maps. The maps are often at regional level and include coordendas geographic meridians and parallels , so we can use them as control points in the georeferencing . Other interesting points to take them as checkpoints are landforms like capes .

There are several methods, but there are two main principles:

1. The two sets of maps must be in the same projection

2. You ahve to select a set of ground control points (like when you are georeferencing)

Good luck :)

If you know the projection of the historical map, then all you need is to change it to match the projection of your modern geodata, most GIS can do this even on the fly. If you do not know the projection, then you need to use control points as suggested in my previous message.

For the country of interest, check out what I have written concerning the Grids and Datums of that nation's geodetic/cartographic history in Photogrammetric Engineering and Remote Sensing journal at www.ASPRS.org.  Once you have obtained that critical information (available free to the public in PDF files), then consult Dr. Gabor Timar's publications on exactly how he has accomplished that task for over ten years in Budapest.  Dr. Timar's work is largely available through Researchgate.net as well as mine for pubs older than 10 years.  (Newer stuff of mine is only available from the ASPRS website.)

In addition to what has already been said about georeferencing, note that the Ground Control Points (GCPs), i.e. the set of homologous points you have to collect on the digitized image of the historical map and the current cartography, should be chosen as points which have preserved unchanged over time. As an example, if you can recognize the same building, pick up a points on its corners. According to the accuracy you want your final product to have, as current cartography don't use a raster map or a Google image (which can have large errors), but try to see if an official vector cartography (possibly at a large scale) is available.

After collecting GCPs, you have to choose a suitable georeferencing model. There are many models that you can choose from (and that you will usually find implemented in GIS software). The results of the transformation can be very different so be careful at which model you choose. To evaluate the georeferencing result, you should use some of the points you have collected as Check Points (CPs), i.e. points not used to compute the model but to validate it. A typically used statistics to evaluate the georeferencing result is the Root Mean Square Error (RMSE).

About the software usable to perform georeferencing, I can suggest the QGIS Georeferencer plugin (which however has some limitations) and the photogrammetric software PCI Orthoengine. The first is open source and available for free, the second is proprietary and available by paying the license.

I suggest two publications that better explain all my points and give you a practical example of georeferencing:

1) Brovelli M. A. and Minghini M. (2012) Georeferencing old maps: a polynomial-based approach for Como historical cadastres. e-Perimetron 7(3), pp. 99-110.2) Brovelli M. A., Minghini M., Giori G. and Beretta M. (2012) Web geoservices and ancient cadastral maps: the Web C.A.R.T.E. project. Transactions in GIS 16(2), pp. 125-142, DOI: 10.1111/j.1467-9671.2012.01311.x

you have to reproject and bring all the historical images into a particular projection (country specific projection system and datum give more accurate in this context) system. Then, you can try to the time series analysis.

In some cases researcher are bring all the images in UTM/WGS1984 projection system.

Whether Historical maps have coordinates or not... in such cases one can use geographical identifiers to match with the current maps....

Apart of the other accuracies to be kept in mind (as already given in previous posts)... strategies always depends upon the type of historical maps and its spatial resolution

After scanning the historical maps, you have to use geographical points that are also represented in the contemporary georeferenced maps to match them, using the Georeferencing editor tool in ArcMap (for instance). This way you are projecting the historical imagens into a particular projection system and datum given by the contemporary map. See and example here: https://www.youtube.com/watch?v=PHtxbpboDro

You have to first identify at least 4 points on the old historical map those you can match with the existing map. These 4 points should be uniformly located throughout the map (not to be concentrated in one corner of the map). Then using these points as "control points" (GCP: Ground Control Points) you can georefrence the old map with the existing one. This procedure is called "rubber sheeting". After matching these 4 control points you will notice that both of the maps exactly overlaps on each other.