This preview has intentionally blurred sections. Sign up to view the full version.
View Full DocumentThis preview has intentionally blurred sections. Sign up to view the full version.
View Full Document
Unformatted text preview: CSE5120fall2009 Indexing Multimedia Databases ? ? ? Problem definition : Multimedia System: A system that can store and retrieve objects/documents with text, voice, im ages, animation, slides show etc. VOICE Query by content : Query based on features in the data content and not on a name or coding of the data file. E.g. query for a picture with a lot of red color, instead of a picture named ”XY0132”. • efficient and • complete (no false dismissals) Sample queries • primary key Find the employee record with emp = 123. • secondary key Find the employee records with salary = 40K. • text Find the documents containing the words mul timedia, indexing . • 1d signals Find companies with similar growth patterns. Find solar wind activity similar to today’s • 2d signals Find photos that look like a sunset over the ocean. Find Xrays with a blob with tumorlike tex ture. Find areas with cornfield colors • 3d signals Find avg. brain scan of epileptics. • Spatial Queries Find the nearest restaurant from a given point Find all movie theaters within 5 miles from CUHK Find nearest restaurant with price &lt; $15 for Won Ton Noodle • Video Find the moment of the end of a horse race. 36 Find the nearest restaurant from a given point: x x x x x x x x x Query (1,1) (5,4) Point P Sequential search: Calculate distance p (5 1) 2 + (4 1) 2 Return point with minimum distance Too slow! What did we do with conventional data search? B + tree 11 19 7 7 10 11 13 19 20 31 52 3 1 14 50 14 31 50 O(log n) If the question was: Is there a restaurant at a (5,4)? Yes we can do it. For example, if the data points are: (1,1), (1,9) (2,1), (3,1), (3,5), (4,0), (4,2), (4,9), (5,4), (6,2), (6,6), (6,9) (1,1) (1,9) (3,1) (4,2) (4,2) (4,9) (6,2) (6,2) (5,4) (6,6) (6,6) (6,9) (4,9) (3,5) (3,5) (4,0) (2,1) (2,1) x x x x x x x x x x ( 5 , 4 ) ( 1 , 5 ) x x Is there a restaurant at a (5,4)? (5,5)? Find the nearest restaurant to (1,5)? Near in first dimension, Far in second dimension. Find all restaurants within 3 km of (1,5) If we live in a one dimensional world: 1 3 6 7 10 11 13 14 15 18 x x x x x x x x x x 3 1 6 18 18 15 14 13 13 11 10 7 11 7 6 15 Find nearest restaurant to (11) Find nearest restaurant to (2). Find nearest restaurant to (17) At most 2 leaf nodes are searched. Find all restaurants within 2 km of (2). O(log N) Observation : 1. Mapping of data to multidimensional points. 2. Query: find near objects 3. Difficulty in handling such querying. 37 kd Trees a b c d e f g h a b c d e f g h Nearest Neighbor query a b c d e f g h Q a b c d e f g h Range query a b c d e f g h Q a b c d e f g h 38 Indexing Multimedia Databases (Continue) For the spatial queries, e.g. find the nearest restaurant to a point, we find the restaurant whose Euclidean distances to the query point is the smallest....
View
Full Document
 Fall '09
 AdaFu
 Databases, Distance, Metric space, Euclidean space, Nearest neighbor search, nearest restaurant

Click to edit the document details