IMatch provides a number of search features that let you find files in your database by various criteria.
IMatch provides a number of search features that let you find files in your database by various criteria. This includes searching for metadata, finding duplicates, using shape/color-based search, and searching for visually similar images.
This is probably the most commonly used search feature. It enables you to search the contents of the File Window (scope) or the entire database for file names, metadata, and attributes; with many options. It works similar to the search bar of your favorite search engine.
See the Search Bar topic in the File Window help for detailed information.
The Quick Filter lives in a separate ribbon above the File Window. It gives quick access to he most-frequently used filters. You toggle it with the ribbon button highlighted in orange on the right:
See the Quick Filter Ribbon help topic for more information.
You use the Filter Panel it to filter the files you see in the File Window.
If offers a wide range of specialized filters, e.g., rating, label, collections, file names, file properties, folders, categories, date & time, metadata tag values, events, people, GPS coordinates, and much more.
See Filter Panel for detailed information.
To access all features, open the Search menu.
IMatch's semantic search is an AI-based tool that uses metadata you have entered manually or created with AutoTagger to improve searches in your database.
See Semantic Search for more information.
Sometimes you end up with duplicate files on your hard drive, e.g. when you accidentally copy the same set of images into different folders or when you download the same files multiple times from your smartphone or cards. IMatch can help you reclaim disk space by finding duplicate files with the Search > Duplicates and Search > Copies commands.
Whether you search for Copies or Duplicates depends on what you want to find:
Duplicates | IMatch considers files as duplicates when the image data is identical or nearly identical. This mode does not consider metadata or other data in the file. It uses an image fingerprint calculated by an algorithm that IMatch stores for each image in the database. Use this to find identical files even after changing and writing-back metadata. If the image data has changed substantially, for example because it was resized, cropped, or modified in other ways, the image will no longer be considered a duplicate. A few changed pixels usually do not cause the algorithm to miss a match, though. This search operation is tuned to find files where the image data is identical or virtually identical to one or more other files. It does not deal with visually similar files. You can use a Visual Query to find visually similar files, e.g. all files with a red object in the center. Converting images between a lossless and a lossy file format (e.g. JPG), or converting between two lossy file formats, does not produce duplicates in the sense used here. The pixel data can change considerably if you convert a PNG file into a JPG file, for instance. Use a Visual Query in such scenarios. |
Duplicates (same format) | Like Duplicates above, but limiting the search to files with the same file format as the original. |
Copies | Search for 100% identical files (bit-by-bit binary identical). Use this mode if you want to identify exact copies of files, e.g. during a clean-up or de-duplication process. |
If you make a copy of beach.jpg to beach-1.jpg in IMatch or Windows Explorer, both files are binary identical. The files will be identified as a duplicate by both search operations. If you now modify metadata for beach.jpg and write-back, the files on disk are no longer binary identical. They look the same, but beach.jpg has other metadata and probably also a different size on disk than beach-1.jpg. The Copies search will no longer consider both files as duplicates.
To use these commands, select the images for which you want to find duplicates in the File Window, then run Search > Copies or Search > Duplicates.
When the search is complete, IMatch opens a Result Window showing the original files (the files selected in the File Window when you used the command) and the duplicates found for each original. Originals without duplicates are removed from the result.
See the Result Windows topic for more info about how IMatch displays search results.
If an original is a match for another original, the original is omitted from the search result. This can happen if you have multiple copies of the same file in the File Window and you select them all for finding copies or duplicates. This prevents you from accidentally deleting every copy of a file because the file is listed as both an original and a match. You can safely press Ctrl+A in the result window to select all duplicates found, and then delete them or run another command.
For example, you select all files in a folder to look for duplicate files. This folder contains two identical files: A and B. B is thus a match for A, and A is a match for B. The result produced by IMatch will only include file A as an original, with B as its match. B will not be listed as an original in the search result.
Press Ctrl+A in the result window to select all duplicates.
The standard File Window layout Result Window for Duplicates uses the special File Window layout attributes:
to provide the information you need to quickly decide which file to keep when searching for duplicates.
The new attributes compare the file size, date & time, and width/height of matches with the corresponding properties of the original file. They are filled for all search results that are based on original files (e.g. duplicate or copy search, visually similar images, ...), making it much easier to decide which duplicate to keep.
The special File Window layout for duplicates provides extra information helpful for deciding which file to keep.
Use the Visually Similar sort profile to sort similar files next to each other.
This helps you visually identify and weed out duplicate files quickly.
You can select all files in the result window, except the originals, with a Ctrl+A and then mark them as rejects with Del or delete them immediately with Ctrl+Del.
This only works when the originals are not displayed alongside their matches. See Result Windows for more information.
If you want IMatch to automatically assign duplicate files or copies to a category (e.g., a category named Dupes), you can enable the corresponding option under Edit > Preferences > Indexing. Enable this before you add the folders containing the duplicates to your database. This makes it easy to spot files that were accidentally added multiple times. Just check the files in the category.
See Indexing for more information.
These search functions look at the actual image content in order to find images with similar colors and shapes.
IMatch matching is not an exact science because what is easy for humans is very, very complicated for computers.
Newer algorithms are developed every year, and IMatch is prepared to incorporate recent developments.
Nevertheless, the visual query features in IMatch can be a big help when answering questions like:
The algorithms used in IMatch are a trade-off between ease of use, performance, and quality. They are designed to help you manage your image collections by allowing you to find duplicate images even if they have been resized, slightly cropped, or modified. They can also help you find images whose colors match a given image for a graphic project or collage, or find all images with certain colors in specific spots.
There may be a short delay when you run the first visual query in an IMatch session.
IMatch loads and caches the visual query base data in memory when it is first accessed. The data is then kept in memory and reused for future visual queries, which run much faster. If you add or update files in your database, IMatch will reload the query data automatically on the next access.
This search module uses information calculated by IMatch when an image is added to the database. It tries to identify images by their dominant colors (the colors used most in the image) and by the overall color distribution in the image. The results are usually quite good, and the performance is excellent even for large databases with 100,000 or more images.
To find visually similar images, first select the files you want to use as the originals in a File Window and then run the Search > Visually similar images... command. IMatch will run the search and open a result window to present the findings.
The settings dialog box for visual queries
In this dialog box you can choose the search type you want to use and control the Shape Importance setting for the shape and color layout search type. If you use a higher value for shape importance IMatch is more tolerant of colors in the image. If you want to give more weight to the colors in the image, use a lower Shape Importance setting.
The Default sort profile in Result Windows sorts by the similarity value resulting from this search operation. Sometimes it is useful to use the Visually Similar sort profile in a Result Window showing the results of a visual query.
The option Favor image with similar dimension is useful when you want to find images from the same set or series, distributed across your entire database. Since images from the same set often have similar dimensions (width/height), this option makes it easier to find images from the same set, even when there are many other files that are visually similar to the original.
Usually a medium or higher value for Shape Importance yields better results. There is no best value for this setting. The effect depends mostly on the type of images you work with and the type of result you expect. Try a few settings to produce the best possible results for a given set of original images.
A typical result for a visual query. Objects with similar colors and shapes were identified and returned in the result.
Searching for bridal bouquets or images with a similar cluster of colors in the center of the image.
(150,000 images in the test database, search time less than 1 second).
Searching by visual similarity will always produce some mismatches, such as the woman or the young family, because it only searches based on color layout and texture. To search by actual image content, use AutoTagger to create keywords based on image content and then search those keywords to, for example, find all images containing a pizza or a cheesecake.
This search type has no additional options. It compares the dominant colors in the original images with the dominant colors of each other image in the database. The best matches are returned in a result window.
This setting allows you to put a higher weight on images with the same orientation (portrait, landscape) as the original image. In many cases you can produce better results when you use this option.
This setting favors images that are roughly the same size as the original image.
When you use the face detection and recognition features, IMatch can use the face data to improve the search results.
If this option is enabled, IMatch performs the visual search as configured. It then checks for faces in the originals and matches. If a match shows the same person as the original, has similar faces, or has faces in roughly the same location, the match gets promoted a bit and moved upwards in the result set.
This option limits the search to files in the current scope instead of searching the entire database. Use this if you want to limit the search to a specific set of images.
If you run a visual query from a Result Window and this option is enabled, the results of the visual query replace the contents of the result window. This makes drilling down into search results more effective.
This setting lets you control how many matches IMatch should show in the result window per original image. Usually a setting of 50 to 200 is best.
If you use many originals (e.g. searching similar images for all images in a folder or category) the result set can become quite large.
For example, if you have 100 originals and you show 100 matches per original, the result window will contain 10,000 files!
This setting controls the minimum similarity of an image to be considered a match. Use this to trim irrelevant matches from the results. Results with similarities under ~80% are rarely relevant.
Since the Result Window suppresses originals without at least one match, this reduces the results you have to deal with when you e.g. process entire folders to find duplicates during a cleanup step.
This visual query feature lets you draw a sketch of the image you're looking for and then have IMatch search the database for all images similar to that sketch. This is very similar to the normal visual query feature, but instead of selecting an existing image as the original for the search, you draw a sketch by hand.
Run Search > Sketch Match to open the Sketch Match dialog box. Fill the grid with the color of your choice by left-clicking and dragging the mouse. Right-click to erase color segments in the grid. IMatch only considers filled grid segments in the search; everything else is considered as "don't care".
Looking for images with blue color at the top.
If you fill a couple of rows at the top of the grid with blue color, you can find all images with blue color in the same area of the image, e.g. images showing a blue sky or another blue object occupying the top of the image. To find all images with a red color in the center, fill a few segments in the middle of the grid with one of the red colors.
Don't try to be too precise. The result will always be a bit fuzzy and a rough sketch is usually good enough to get a usable result.
The number of results can be controlled via the slider in the dialog box.
If you are not satisfied with the result, open the dialog box again via Ctrl+M,K and change the color or position of the filled segments, then repeat the search. IMatch sends the result to a new result window so you can easily compare the individual results.
Result windows are normal File Windows. You can use the filter panel to further "drill down" the results produced by Sketch Match or any of the other search functions.
| Ctrl+M,V | Runs a visual query and opens the settings dialog box. |
| Ctrl+M,M | Runs the last-used visual query again with the currently selected images. This command allows you to drill down quickly in search results. If you spot an interesting image in a result or File Window, just press Ctrl+M,M again to use this image as the original for a new search. |
| Ctrl+M,K | Runs a sketch match and opens the Sketch Match dialog box. |
Press Ctrl+M,M to rerun the visual query with the last-used settings, or choose Rerun from the Search menu.
Please note that visual queries do not work very well with images that have transparencies or alpha channels. If you search for partially transparent images, the quality of the result will vary considerably, depending on the other contents of your database.
This search feature lets you find files nearby the currently selected files or files at specific locations.
This feature uses the GPS coordinates IMatch extracts from your files. Files without GPS coordinates are ignored by this feature.
The GPS Search
To find all files within a given maximum distance for the currently selected files, enable the Find files nearby mode. Specify a maximum distance to define the search radius. IMatch uses the coordinate of the selected file and this distance to determine nearby files. If you also want to take altitude into account, enable this option and specify the maximum altitude difference to consider during the search.
If you enable this search mode, you can use the locations you have created in the GPS panel to find all files taken at one or more locations. You can either specify a maximum distance and altitude difference, or use the map radius specified in the location to find files at that location.
You can search for files within a specific region using features provided by the Map Panel.