Search Patterns In Art

This app search patterns in painting images using different computer vision techniques (SIFT/SURF, Kmeans, RANSAC, Homography).

The program reads a set of images predefined by the user. These images contains some patterns (objects) that we want to compare tith other new image . These represent a learning image set or a "vocabulary". Every image of this vocabulary represents a "word". This words will be used by the program to recognize other possible similar patterns in a new image.

Installation

You need:

• C++ Compiler
• OpenCV 2.4.3 library (http://opencv.willowgarage.com/wiki/).

License

This program is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version.

This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.

You should have received a copy of the GNU General Public License along with this program. If not, see http://www.gnu.org/licenses/.

Usage

Once you have downloaded the source code you have to define some CONSTANTS to run the application. These are defined in searchPatterns.cpp:

• algorithmType (For example = "SIFT"): The detector keypoints and type. This can be FAST, STAR, SIFT, SURF, ORB, MSER, GFTT, HARRIS, Dense, SimpleBlob ...

Only for SURF algorithmType:

• uprightSURF : This is USURF. false=detector computes orientation of each feature. true= the orientation is not computed.
• hessianThresholdSURF : Threshold for the keypoint detector. A good default value could be from 300 to 500, depending from the image contrast.
• nOctaves : Number of pyramid octaves the keypoint detector will use.
• nOctaveLayers : Number of octave layers within each octave.
• extended : Extended descriptor flag (true - use extended 128-element descriptors; false - use 64-element descriptors).

Image Effects (Gaussian Blur, resize):

• kernelSize : This means the Gaussian kernel size applied to newImage. (-1: Not apply)
• resizeImage : This means if we make a resize transformation of the image

K-Means:

• initialK : Initial K Center constant in k-means. This must be <= Total number of rows in the sum of all vocabulary images.
• kIncrement : This is the increment of the k centers in kmeans loop
• criteriaKMeans : This is the maximum number of iterations in kmeans to recalcule the k-centers (Ex: 100 it's ok)
• attemptsKMeans : This is the number of times the algorithm is executed using different initial labellings (Ex: 3 it's ok)

RANSAC:

• minimumVotes : Minimum number of votes that must to have every image to be selected. (Minimum 2.Homography needs 2 points minimum) (Ex: 8-10 are good values)
• thresholdDistanceAdmitted : Threshold distance admitted comparing distance between images on homography results. (Ex: 30 it's ok)
• homographyAttempts : Number of RANSAC attempts to find homographies

Directories, files:

• vocabularyImagesNameFile (For example = "/../vocabularyImages.txt") This .txt file contains the name of the learning image set Every image represents a "word" inside the "vocabulary" of learning image set
• newImageFileName (For example = "/../tapies9.jpg"): This is the new image that we want to compare with the learning image set
• dirToSaveResImages (For example = "/../results"): This is the route/directory of to save the result images