/*
* This file is part of the LIRE project: http://lire-project.net
* LIRE 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 2 of the License, or
* (at your option) any later version.
*
* LIRE 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 LIRE; if not, write to the Free Software
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
*
* We kindly ask you to refer the any or one of the following publications in
* any publication mentioning or employing Lire:
*
* Lux Mathias, Savvas A. Chatzichristofis. Lire: Lucene Image Retrieval –
* An Extensible Java CBIR Library. In proceedings of the 16th ACM International
* Conference on Multimedia, pp. 1085-1088, Vancouver, Canada, 2008
* URL: http://doi.acm.org/10.1145/1459359.1459577
*
* Lux Mathias. Content Based Image Retrieval with LIRE. In proceedings of the
* 19th ACM International Conference on Multimedia, pp. 735-738, Scottsdale,
* Arizona, USA, 2011
* URL: http://dl.acm.org/citation.cfm?id=2072432
*
* Mathias Lux, Oge Marques. Visual Information Retrieval using Java and LIRE
* Morgan & Claypool, 2013
* URL: http://www.morganclaypool.com/doi/abs/10.2200/S00468ED1V01Y201301ICR025
*
* Copyright statement:
* ====================
* (c) 2002-2013 by Mathias Lux (mathias@juggle.at)
* http://www.semanticmetadata.net/lire, http://www.lire-project.net
*
* Updated: 13.02.15 19:17
*/
package net.semanticmetadata.lire.searchers.custom;
import net.semanticmetadata.lire.imageanalysis.features.GlobalFeature;
import net.semanticmetadata.lire.imageanalysis.features.global.CEDD;
import net.semanticmetadata.lire.searchers.*;
import net.semanticmetadata.lire.utils.MetricsUtils;
import org.apache.lucene.document.Document;
import org.apache.lucene.index.IndexReader;
import java.awt.image.BufferedImage;
import java.io.IOException;
import java.util.*;
import java.util.logging.Level;
import java.util.logging.Logger;
/**
* A ImageSearcher that retrieves just the first result, caches the whole index and optimizes search time by
* bundling searches. Please note that as soon as the instance is created, changes in the index are not
* reflected.
*
* @author Mathias Lux, mathias@juggle.at
*/
public class SingleNddCeddImageSearcher extends AbstractImageSearcher {
protected Logger logger = Logger.getLogger(getClass().getName());
Class<?> descriptorClass = CEDD.class;
String fieldName = null;
protected GlobalFeature cachedInstance = null;
protected boolean isCaching = true;
protected ArrayList<double[]> featureCache;
protected IndexReader reader;
protected TreeSet<SimpleResult> docs;
HashMap<double[], LinkedList<Integer>> hashMap;
protected double maxDistance;
protected boolean useSimilarityScore = false;
private boolean halfDimensions = false;
/**
* Creates a new ImageSearcher for searching just one single image based on CEDD from a RAM cached data set.
*
* @param reader the index reader pointing to the index. It will be cached first, so changes will not be reflected in this instance.
*/
public SingleNddCeddImageSearcher(IndexReader reader) {
init(reader);
}
/**
* Creates a new ImageSearcher for searching just one single image based on CEDD from a RAM cached data set.
* Set approximate to true if you want to speed up search and loose accuracy.
*
* @param reader the index reader pointing to the index. It will be cached first, so changes will not be reflected in this instance.
* @param approximate set to true if you want to trade accuracy to speed, setting to true is faster (~ double speed), but less accurate
*/
public SingleNddCeddImageSearcher(IndexReader reader, boolean approximate) {
this.halfDimensions = approximate;
init(reader);
}
/**
* Eventually to be used with other LireFeature classes.
* @param reader
* @param approximate
* @param descriptorClass
*/
public SingleNddCeddImageSearcher(IndexReader reader, boolean approximate, Class descriptorClass, String fieldName) {
this.halfDimensions = approximate;
this.descriptorClass = descriptorClass;
this.fieldName = fieldName;
init(reader);
}
protected void init(IndexReader reader) {
this.reader = reader;
if (reader.hasDeletions()) {
throw new UnsupportedOperationException("The index has to be optimized first to be cached! Use IndexWriter.forceMerge(0) to do this.");
}
docs = new TreeSet<SimpleResult>();
try {
this.cachedInstance = (GlobalFeature) this.descriptorClass.newInstance();
if (fieldName == null) fieldName = this.cachedInstance.getFieldName();
} catch (InstantiationException e) {
logger.log(Level.SEVERE, "Error instantiating class for generic image searcher (" + descriptorClass.getName() + "): " + e.getMessage());
} catch (IllegalAccessException e) {
logger.log(Level.SEVERE, "Error instantiating class for generic image searcher (" + descriptorClass.getName() + "): " + e.getMessage());
}
// put all respective features into an in-memory cache ...
if (isCaching && reader != null) {
int docs = reader.numDocs();
featureCache = new ArrayList<double[]>(docs);
try {
Document d;
for (int i = 0; i < docs; i++) {
d = reader.document(i);
cachedInstance.setByteArrayRepresentation(d.getField(fieldName).binaryValue().bytes, d.getField(fieldName).binaryValue().offset, d.getField(fieldName).binaryValue().length);
// normalize features,o we can use L1
if (!halfDimensions) {
featureCache.add(normalize(cachedInstance.getFeatureVector()));
} else {
featureCache.add(crunch(cachedInstance.getFeatureVector()));
}
}
} catch (IOException e) {
e.printStackTrace();
}
}
}
private double[] normalize(double[] doubleHistogram) {
double[] result = new double[doubleHistogram.length];
for (int i = 0; i < doubleHistogram.length; i++) {
result[i] = doubleHistogram[i] / 8d;
}
return result;
}
/**
* Reduces dimensions of CEDD to half while normalizing the vector.
* @param doubleHistogram
* @return
*/
private double[] crunch(double[] doubleHistogram) {
double[] result = new double[doubleHistogram.length / 2];
for (int i = 0; i < doubleHistogram.length; i += 2) {
result[i / 2] = doubleHistogram[i] + doubleHistogram[i + 1] / 16d;
}
return result;
}
public ImageSearchHits search(BufferedImage image, IndexReader reader) throws IOException {
throw new UnsupportedOperationException("not implemented!");
}
/**
* @param reader
* @param globalFeature
* @return the maximum distance found for normalizing.
* @throws java.io.IOException
*/
protected double findSimilar(IndexReader reader, GlobalFeature globalFeature) throws IOException {
maxDistance = -1;
// clear result set ...
docs.clear();
double tmpDistance;
// we use the in-memory cache to find the matching docs from the index.
int count = 0;
double[] doubleHistogram;
if (!halfDimensions) {
doubleHistogram = normalize(globalFeature.getFeatureVector());
} else {
doubleHistogram = crunch(globalFeature.getFeatureVector());
}
double[] tmp;
int index = -1;
for (Iterator<double[]> iterator = featureCache.iterator(); iterator.hasNext(); ) {
tmp = iterator.next();
tmpDistance = MetricsUtils.distL1(doubleHistogram, tmp);
assert (tmpDistance >= 0);
if (tmpDistance < maxDistance) {
maxDistance = tmpDistance;
index = count;
}
count++;
}
this.docs.add(new SimpleResult(maxDistance, index));
return maxDistance;
}
public SimpleResult findMostSimilar(GlobalFeature globalFeature) throws IOException {
findSimilar(reader, globalFeature);
return docs.first();
}
public SimpleResult[] findMostSimilar(GlobalFeature[] globalFeatures) throws IOException {
return findMostSimilar(globalFeatures, 0, globalFeatures.length);
}
public SimpleResult[] findMostSimilar(GlobalFeature[] globalFeatures, int offset, int length) throws IOException {
double[] maxDistanceArray = new double[length - offset];
Arrays.fill(maxDistanceArray, Double.MAX_VALUE);
double tmpDistance;
int count = 0;
double[][] dhs = new double[0][];
try {
dhs = new double[length][featureCache.get(0).length];
} catch (Exception e) {
e.printStackTrace();
}
for (int i = 0; i < dhs.length; i++) {
if (!halfDimensions) {
dhs[i] = normalize(globalFeatures[offset + i].getFeatureVector());
} else {
dhs[i] = crunch(globalFeatures[offset + i].getFeatureVector());
}
}
double[] tmp;
int[] indexes = new int[length];
Arrays.fill(indexes, -1);
for (Iterator<double[]> iterator = featureCache.iterator(); iterator.hasNext(); ) {
tmp = iterator.next();
for (int i = 0; i < dhs.length; i++) {
tmpDistance = MetricsUtils.distL1(dhs[i], tmp);
assert (tmpDistance >= 0);
if (tmpDistance < maxDistanceArray[i]) {
maxDistanceArray[i] = tmpDistance;
indexes[i] = count;
}
}
count++;
}
SimpleResult[] results = new SimpleResult[length];
for (int i = 0; i < results.length; i++) {
if (indexes[i] >= 0 && indexes[i] < reader.maxDoc())
results[i] = new SimpleResult(maxDistanceArray[i], indexes[i]);
else
results[i] = null;
}
return results;
}
/**
* Main similarity method called for each and every document in the index.
*
* @param document
* @param globalFeature
* @return the distance between the given feature and the feature stored in the document.
*/
protected double getDistance(Document document, GlobalFeature globalFeature) {
if (document.getField(fieldName).binaryValue() != null && document.getField(fieldName).binaryValue().length > 0) {
cachedInstance.setByteArrayRepresentation(document.getField(fieldName).binaryValue().bytes, document.getField(fieldName).binaryValue().offset, document.getField(fieldName).binaryValue().length);
return globalFeature.getDistance(cachedInstance);
} else {
logger.warning("No feature stored in this document! (" + descriptorClass.getName() + ")");
}
return 0d;
}
public ImageSearchHits search(Document doc, IndexReader reader) throws IOException {
SimpleImageSearchHits searchHits = null;
try {
GlobalFeature globalFeature = (GlobalFeature) descriptorClass.newInstance();
if (doc.getField(fieldName).binaryValue() != null && doc.getField(fieldName).binaryValue().length > 0)
globalFeature.setByteArrayRepresentation(doc.getField(fieldName).binaryValue().bytes, doc.getField(fieldName).binaryValue().offset, doc.getField(fieldName).binaryValue().length);
double maxDistance = findSimilar(reader, globalFeature);
if (!useSimilarityScore) {
searchHits = new SimpleImageSearchHits(this.docs, maxDistance);
} else {
searchHits = new SimpleImageSearchHits(this.docs, maxDistance, useSimilarityScore);
}
} catch (InstantiationException e) {
logger.log(Level.SEVERE, "Error instantiating class for generic image searcher: " + e.getMessage());
} catch (IllegalAccessException e) {
logger.log(Level.SEVERE, "Error instantiating class for generic image searcher: " + e.getMessage());
}
return searchHits;
}
public ImageDuplicates findDuplicates(IndexReader reader) throws IOException {
throw new UnsupportedOperationException("not implemented!");
}
public String toString() {
return "GenericSearcher using " + descriptorClass.getName();
}
}