I need to extract nice land use classification from old panchromatic photos 1954 and 1968 with quite good resolution: 0,96 m. I am not an expert in image processing of panchromatic ones. I know that I will build a model to test the quality of image enhancement many times, but at the moment I need some advice in:
1) how to get rid of the shadows or at least decrease their effect? (ERDAS 2013, ARCGIS etc.) 2) how to extract water bodies from PAN? I work with coastal areas with many inland water bodies... 3) any suggestions for a urban - non urban classification? I was thinking about combining results of textural and spectral enhancement (new channel from combining the TEXTURE AND CONVULTION in ERDAS 2013) and setting a treshold in the model maker
Regarding the shadows, you may remove part of it using saturation (HSV) filters (resampling or filtering tools), at the cost of a quality loss. This can be done uneder ENVI, ArcGIS or even photo processing softwares. For example try to play with colors and saturation under The Gimp (free) or any photoshop-like if you have some.
The other points (water plans, urban areas) can be partially solved (especially the waterplans) with assisted classification tools (ENVI of course and some tool in some arcgis extension, maybe spatial analyst, does it) BUT there is no absolute magic, and you are working on a panchromatic imagery so discrimination is harder than with multispectral datasets.
The last -and ultimate- classification tool, is human eye coupled with an active brain and high speed mouse clicking fingers so you can digitise recognisable features. And I assume that on a 1m resolution air photo anthropic activity must be easy to spot.
You may also use some topographic data to generate your watershed and coast line topo break. some nice datasets are sometimes freely available depending on your place or your status (and researcher is good!).