Main Content

retrieveImages

Search image set for similar image

Description

example

imageIDs = retrieveImages(queryImage,imageIndex) returns the image identifiers imageIDs that correspond to images within imageIndex that are visually similar to the query image. The imageIDs are returned in rankeded order, from the most to least similar match.

example

[imageIDs,scores] = retrieveImages(queryImage,imageIndex) optionally returns the similarity scores used to rank the image retrieval results. The scores output contains the corresponding scores from 0 to 1.

example

[imageIDs,scores,imageWords] = retrieveImages(queryImage,imageIndex) optionally returns the visual words in queryImage that are used to search for similar images.

example

[imageIDs,___] = retrieveImages(queryImage,imageIndex,Name,Value) uses additional options specified by one or more Name,Value pair arguments, using any of the preceding syntaxes.

Examples

collapse all

Create an image set of book covers.

dataDir = fullfile(toolboxdir('vision'),'visiondata','bookCovers');
bookCovers = imageDatastore(dataDir);

Display the data set.

thumbnailGallery = [];
for i = 1:length(bookCovers.Files)
    I = readimage(bookCovers,i);
    thumbnail = imresize(I,[300 300]);
    thumbnailGallery = cat(4,thumbnailGallery,thumbnail);
end

figure
montage(thumbnailGallery);

Figure contains an axes object. The axes object contains an object of type image.

Index the image set. This step may take a few minutes.

imageIndex = indexImages(bookCovers);
Creating an inverted image index using Bag-Of-Features.
-------------------------------------------------------

Creating Bag-Of-Features.
-------------------------

* Selecting feature point locations using the Detector method.
* Extracting SURF features from the selected feature point locations.
** detectSURFFeatures is used to detect key points for feature extraction.

* Extracting features from 58 images...done. Extracted 29216 features.

* Keeping 80 percent of the strongest features from each category.

* Balancing the number of features across all image categories to improve clustering.
** Image category 1 has the least number of strongest features: 23373.
** Using the strongest 23373 features from each of the other image categories.

* Creating a 20000 word visual vocabulary.
* Number of levels: 1
* Branching factor: 20000
* Number of clustering steps: 1

* [Step 1/1] Clustering vocabulary level 1.
* Number of features          : 23373
* Number of clusters          : 20000
* Initializing cluster centers...100.00%.
* Clustering...completed 7/100 iterations (~0.58 seconds/iteration)...converged in 7 iterations.

* Finished creating Bag-Of-Features


Encoding images using Bag-Of-Features.
--------------------------------------

* Encoding 58 images...done.
Finished creating the image index.

Select and display the query image.

queryDir = fullfile(dataDir,'queries',filesep);
queryImage = imread([queryDir 'query3.jpg']);

imageIDs = retrieveImages(queryImage,imageIndex);

Show the query image and its best match, side-by-side.

bestMatch = imageIDs(1);
bestImage = imread(imageIndex.ImageLocation{bestMatch});

figure
imshowpair(queryImage,bestImage,'montage')

Figure contains an axes object. The axes object contains an object of type image.

Search an image set for an object using a region of interest (ROI) for the query image.

Define a set of images to search.

imageFiles = ...
  {'elephant.jpg', 'cameraman.tif', ...
  'peppers.png',  'saturn.png',...
  'pears.png',    'stapleRemover.jpg', ...
  'football.jpg', 'mandi.tif',...
  'kids.tif',     'liftingbody.png', ...
  'office_5.jpg', 'gantrycrane.png',...
  'moon.tif',     'circuit.tif', ...
  'tape.png',     'coins.png'};

imds = imageDatastore(imageFiles);

Create a search index.

 imageIndex = indexImages(imds);
Creating an inverted image index using Bag-Of-Features.
-------------------------------------------------------

Creating Bag-Of-Features.
-------------------------

* Selecting feature point locations using the Detector method.
* Extracting SURF features from the selected feature point locations.
** detectSURFFeatures is used to detect key points for feature extraction.

* Extracting features from 16 images...done. Extracted 3680 features.

* Keeping 80 percent of the strongest features from each category.

* Balancing the number of features across all image categories to improve clustering.
** Image category 1 has the least number of strongest features: 2944.
** Using the strongest 2944 features from each of the other image categories.

* Creating a 2944 word visual vocabulary.
* Number of levels: 1
* Branching factor: 2944
* Number of clustering steps: 1

* [Step 1/1] Clustering vocabulary level 1.
* Number of features          : 2944
* Number of clusters          : 2944
* Initializing cluster centers...100.00%.
* Clustering...completed 1/100 iterations (~0.08 seconds/iteration)...converged in 1 iterations.

* Finished creating Bag-Of-Features


Encoding images using Bag-Of-Features.
--------------------------------------

* Encoding 16 images...done.
Finished creating the image index.

Specify a query image and an ROI. The ROI outlines the object, an elephant, for the search.

queryImage = imread('clutteredDesk.jpg');
queryROI = [130 175 330 365];

figure
imshow(queryImage)
rectangle('Position',queryROI,'EdgeColor','yellow')

Figure contains an axes object. The axes object contains 2 objects of type image, rectangle.

You can also use the imrect function to select an ROI interactively. For example, queryROI = getPosition(imrect)

Find images that contain the object.

imageIDs = retrieveImages(queryImage,imageIndex,'ROI',queryROI)
imageIDs = 12x1 uint32 column vector

    1
   11
    6
   12
    2
    3
    8
   10
   13
   14
      ⋮

Display the best match.

bestMatch = imageIDs(1);

figure
imshow(imageIndex.ImageLocation{bestMatch})

Figure contains an axes object. The axes object contains an object of type image.

Use the locations of visual words to verify the best search result. To rerank the search results based on geometric information, repeat this procedure for the top N search results.

Specify the location of the images.

dataDir = fullfile(toolboxdir('vision'),'visiondata','bookCovers');
bookCovers = imageDatastore(dataDir);

Index the image set. This process can take a few minutes.

imageIndex = indexImages(bookCovers);
Creating an inverted image index using Bag-Of-Features.
-------------------------------------------------------

Creating Bag-Of-Features.
-------------------------

* Selecting feature point locations using the Detector method.
* Extracting SURF features from the selected feature point locations.
** detectSURFFeatures is used to detect key points for feature extraction.

* Extracting features from 58 images...done. Extracted 29216 features.

* Keeping 80 percent of the strongest features from each category.

* Balancing the number of features across all image categories to improve clustering.
** Image category 1 has the least number of strongest features: 23373.
** Using the strongest 23373 features from each of the other image categories.

* Creating a 20000 word visual vocabulary.
* Number of levels: 1
* Branching factor: 20000
* Number of clustering steps: 1

* [Step 1/1] Clustering vocabulary level 1.
* Number of features          : 23373
* Number of clusters          : 20000
* Initializing cluster centers...100.00%.
* Clustering...completed 7/100 iterations (~0.63 seconds/iteration)...converged in 7 iterations.

* Finished creating Bag-Of-Features


Encoding images using Bag-Of-Features.
--------------------------------------

* Encoding 58 images...done.
Finished creating the image index.

Select and display the query image.

queryDir = fullfile(dataDir,'queries',filesep);
queryImage = imread([queryDir 'query3.jpg']);

figure
imshow(queryImage)

Figure contains an axes object. The axes object contains an object of type image.

Retrieve the best matches. The queryWords output contains visual word locations information for the query image. Use this information to verify the search results.

[imageIDs, ~, queryWords] = retrieveImages(queryImage,imageIndex);

Find the best match for the query image by extracting the visual words from the image index. The image index contains the visual word information for all images in the index.

bestMatch = imageIDs(1);
bestImage = imread(imageIndex.ImageLocation{bestMatch});
bestMatchWords = imageIndex.ImageWords(bestMatch);

Generate a set of tentative matches based on visual word assignments. Each visual word in the query can have multiple matches due to the hard quantization used to assign visual words.

queryWordsIndex     = queryWords.WordIndex;
bestMatchWordIndex  = bestMatchWords.WordIndex;

tentativeMatches = [];
for i = 1:numel(queryWords.WordIndex)
    
    idx = find(queryWordsIndex(i) == bestMatchWordIndex);
    
    matches = [repmat(i, numel(idx), 1) idx];
    
    tentativeMatches = [tentativeMatches; matches];
    
end

Show the point locations for the tentative matches. There are many poor matches.

points1 = queryWords.Location(tentativeMatches(:,1),:);
points2 = bestMatchWords.Location(tentativeMatches(:,2),:);

figure
showMatchedFeatures(queryImage,bestImage,points1,points2,'montage')

Figure contains an axes object. The axes object contains 4 objects of type image, line.

Remove poor visual word assignments using estimateGeometricTransform2D function. Keep the assignments that fit a valid geometric transform.

[tform,inlierIdx] = ...
    estimateGeometricTransform2D(points1,points2,'affine',...
        'MaxNumTrials',2000);
inlierPoints1 = points1(inlierIdx, :);
inlierPoints2 = points2(inlierIdx, :);

Rerank the search results by the percentage of inliers. Do this when the geometric verification procedure is applied to the top N search results. Those images with a higher percentage of inliers are more likely to be relevant.

percentageOfInliers = size(inlierPoints1,1)./size(points1,1);

figure
showMatchedFeatures(queryImage,bestImage,inlierPoints1,...
    inlierPoints2,'montage')

Figure contains an axes object. The axes object contains 4 objects of type image, line.

Apply the estimated transform.

outputView = imref2d(size(bestImage));
Ir = imwarp(queryImage, tform, 'OutputView', outputView);

figure
imshowpair(Ir,bestImage,'montage')

Figure contains an axes object. The axes object contains an object of type image.

Use the evaluateImageRetrieval function to help select proper search parameters.

Create an image set.

setDir  = fullfile(toolboxdir('vision'),'visiondata','imageSets','cups');
imds = imageDatastore(setDir, 'IncludeSubfolders', true, 'LabelSource', 'foldernames');

Index the image set.

 imageIndex = indexImages(imds,'Verbose',false);

Tune image search parameters.

imageIndex.MatchThreshold = 0.2;
imageIndex.WordFrequencyRange = [0 1]
imageIndex = 
  invertedImageIndex with properties:

         ImageLocation: {6x1 cell}
            ImageWords: [6x1 vision.internal.visualWords]
         WordFrequency: [1x1366 double]
         BagOfFeatures: [1x1 bagOfFeatures]
               ImageID: [1 2 3 4 5 6]
        MatchThreshold: 0.2000
    WordFrequencyRange: [0 1]

queryImage = readimage(imds, 1);
indices = retrieveImages(queryImage,imageIndex);

Input Arguments

collapse all

Input query image, specified as either an M-by-N-by-3 truecolor image or an M-by-N 2-D grayscale image.

Data Types: single | double | int16 | uint8 | uint16 | logical

Image search index, specified as an invertedImageIndex object. The indexImages function creates the invertedImageIndex object, which stores the data used for the image search.

Name-Value Arguments

Specify optional comma-separated pairs of Name,Value arguments. Name is the argument name and Value is the corresponding value. Name must appear inside quotes. You can specify several name and value pair arguments in any order as Name1,Value1,...,NameN,ValueN.

Example: 'NumResults',25 sets the 'NumResults' property to 25

Maximum number of results to return, specified as the comma-separated pair consisting of 'NumResults' and a numeric value. Set this value to Inf to return as many matching images as possible.

Query image search region, specified as the comma-separated pair consisting of 'ROI' and an [x y width height] vector.

Similarity metric used to rank the image retrieval results, specified as 'cosine' or 'L1' [3].

Output Arguments

collapse all

Ranked index of retrieved images, returned as an M-by-1 vector. The image IDs are returned in ranked order, from the most to least similar matched image.

Similarity metric, returned as an N-by-1 vector. This output contains the scores that correspond to the retrieved images in the imageIDs output. The scores are computed using the Metric property and a range from 0 to 1.

Object for storing visual word assignments, returned as a visualWords object. The object stores the visual word assignments of queryImage and their locations within that image.

References

[1] Sivic, J. and A. Zisserman. Video Google: A text retrieval approach to object matching in videos. ICCV (2003) pg 1470-1477.

[2] Philbin, J., O. Chum, M. Isard, J. Sivic, and A. Zisserman. Object retrieval with large vocabularies and fast spatial matching. CVPR (2007).

[3] Gálvez-López, Dorian, and Juan D. Tardos. Bags of binary words for fast place recognition in image sequences. IEEE Transactions on Robotics 28.5 (2012): 1188-1197.

Introduced in R2015a