Caliph and Emir

Caliph and Emir are a pair of complementary, open-source Java applications designed for MPEG-7-based photo annotation and Content-Based Image Retrieval (CBIR). Developed by researcher Mathias Lux, they function as a two-part ecosystem:

Caliph (Common And Light-weight PHoto annotation): Used to analyze, tag, and extract technical and semantic data from your images, saving them as structured XML files.

Emir (Experimental Metadata based Image Retrieval): The dedicated search engine that indexes the files created by Caliph so you can perform advanced visual and textual searches. Step 1: Pre-Processing and Annotation (Using Caliph)

Before you can search for any images using Emir, you must first feed your photo repository through Caliph to generate the necessary metadata.

Automatic Feature Extraction: When you load a directory of photos into Caliph, it automatically extracts low-level visual descriptors based on the MPEG-7 standard. These include ColorLayout (spatial distribution of colors), ScalableColor (global color histogram), and EdgeHistogram (spatial distribution of edges/shapes). It also parses existing embedded EXIF and IPTC metadata.

Semantic Graph Modeling: You can manually enrich your photos by drawing semantic graphs. Instead of basic tags, you define real-world objects and their relationships (e.g., “Person A” is next to “Object B” at “Location C”).

Batch Processing: To speed up the process, use Caliph’s batch annotation tool. You can select an entire event folder, trigger the low-level visual feature extraction, and bulk-assign common tags (like date or location) to all photos simultaneously. Step 2: Querying and Image Searching (Using Emir)

Once Caliph outputs the metadata files into your local file directory, open Emir and point it to that directory to utilize its comprehensive search modes:

[ Your Images ] ──> ( CALIPH: Extracts MPEG-7 Features & Graphs ) ──> [ XML Metadata ] │ ┌─────────────────────── EMIR SEARCH ENGINE ───────────────────────────────┘ ▼ ▼ ▼ [ Content-Based Search ] [ Keyword / Text Search ] [ Graph-Based Querying ] - Color Layout - Text Tags - Visualizes “Objects” - Edge Histograms - Lucene-powered - Maps Complex Relations - Query-by-Example - XPath file filtering - Supports Wildcards 1. Content-Based Image Retrieval (CBIR)

This mode lets you search without using words. Emir supports Query-by-Example (QBE). Caliph & Emir: MPEG-7 photo annotation and retrieval