This project consists of four libraries:

• 104Core - basic data structures, like Image, Point, etc.
• 104Algo - basic algorithms, like convolution, centered gradient, etc.
• 104IO - image input-output and some tools for visualization.
• MSAS - stands for "MultiScale Analysis of Similarities" and contains related algorithms.

and two executables:

• StructureTensorApp - computes Affine Covariant Structure Tensors for a given image.
• SimilarityMapApp - computes similarity map for a given point on one image and all points on another image.

Dependencies:

• libpng
• libjpeg
• libtiff

In order to compile everything, create a new folder for build artifacts (e.g. named build) and run cmake and make from there:

mkdir build
cd build
cmake ..
make

Compiled libraries and executables will be placed in the build/bin/ folder.

Note: all commands are ussumed to be executed from the build/bin/ folder.

To see usage information do

StructureTensorApp --help

and

SimilarityMapApp --help

Draw elliptical patched on a set of points specified in the text file 'points.txt':

StructureTensorApp image.png -m ellipses --points points.txt

Set color of elliptical patches:

StructureTensorApp image.png -m ellipses --hue 120

Compute structure tensors using given parameters (number of iterations = 35, radius = 150) and write the result to the file with 'result' prefix:

StructureTensorApp image.png -i 35 -r 150 -o result

Compute average size of elliptical patches on a set of points regularly distributed over the image every 25 pixels:

StructureTensorApp image.png -m avg_size -s 25

Compute similarity map between the point (100, 25) on 'image.png' and whole 'image2.png':

SimilarityMapApp image.png image2.png -p 100:25

Compute similarity map between the point (220, 310) and all other points of the image 'image.png' and write the result to the file with 'result' prefix:

SimilarityMapApp image.png -p 220:310 -o result

Compute similarity map as before but output raw distances in a text file:

SimilarityMapApp image.png -p 220:310 -o result --raw

Compute similarity map as before specifying scale, radius and gamma parameters:

SimilarityMapApp image.png -p 220:310 -t 0.1 -r 150 -v 1.0 -g 0.6