Abstract
With the proliferation of digital cameras and self-publishing of photos, automatic detection of image orientation has become an important part of photo-management systems. In this paper, we present a novel system, based on combining the outputs of hundreds of classifiers trained with AdaBoost, to determine the upright orientation of an image. We thoroughly test our system on photos gathered from professional and amateur photo collections that have been taken with a variety of cameras (digital, film, camera phones). The test images include photos that are in color and black and white, realistic and abstract, and outdoor and indoor. As this system is intended for mass consumer deployment, efficiency in use and accessibility is paramount. Results show that the presented method surpasses similar methods based on Support Vector Machines, in terms of both accuracy and feasibility of deployment.
Similar content being viewed by others
References
Vailaya A, Zhang H, Yang C, Liu F, Jain A (2002) Automatic image orientation detection. IEEE Trans Image Process 11:7
Wang Y, Zhang H (2001) Content-based image orientation detection with support vector machines. In: IEEE workshop on content-based access of image and video libraries, pp 17–23
Wang Y, Zhang H (2004) Detecting image orientation based on low level visual content. Comput Vis Image Underst 93(3):328–346
Zhang, L, Li M, Zhang, H (2002) Boosting image orientation detection with indoor versus outdoor classification. In: Workshop on application of computer vision
Luo J, Boutell M (2005) A probabilistic approach to image orientation detection via confidence-based integration of low level and semantic cues. IEEE Trans Pattern Anal Mach Intell 27(5):715–726
Lyu S (2005) Automatic image orientation determination with natural image statistics. In: Proceedings of the thirteenth annual ACM international conference on multimedia, pp 491–494
Wang L, Liu X, Xia, L, Xu G, Bruckstein A (2003) Image orientation detection with integrated human perception cues (or which way is up). ICIP
Baluja S, Rowley HA (2005) Boosting sex identification performance. In: Proceedings of the twentieth national conference on artificial intelligence and the seventeenth innovative applications of artificial intelligence conference, pp 1508–1513
LeCun Y, Bottou L, Bengio Y, Haffner P (1998) Gradient based learning applied to document recognition. Proc IEEE 86:11
Rowley H, Baluja S, Kanade T (1998) Neural network-based face detection. IEEE Trans Pattern Anal Mach Intell 20(1):23–28
Viola P, Jones MJ (2001) Rapid object detection using a boosted cascade of simple features. In: 2001 IEEE computer society conference on computer vision and pattern recognition (CVPR’01), p 511
Freund Y, Schapire R (1996) Experiments with a new boosting algorithm. In: Machine learning, proceedings of the thirteenth international conference
Wu J, Rehg J, Mullins M (2003) Learning a rare event detection cascade by direct feature selection. Neural Inf Process Syst 16
Cortes C, Vapnik V (1995) Support-vector-networks. Mach Learn 20(3):273–297
Joachims T (1999) Making large-scale SVM learning practical. In: Schölkopf B, Burges C, Smola A (eds) Advances in Kernel methods—support vector learning
Corel-Gallery 1,300,000 (1999) Image gallery—16 CD Set—JB #40629
Shapiro L (2002) University of Washington image database for content-based image retrieval. available from http://www.cs.washington.edu/research/imagedatabase/report03.html
Dietterich TG (2000) Ensemble methods in machine learning. In: Roli I, Kittler J (eds) First international workshop on multiple classifier systems. Lecture notes in computer science. Springer, Berlin Heidelberg New York, pp 1–15
Ghosh J (2002) MultiClassifier systems: back to the future. In: Roli I, Kittler J (eds) Multiple classifier systems. LNCS 2364:1–15
Luo J, Crandall D, Singhal A, Boutell M, Gray R (2003) Psychophysical study of image orientation perception. Spat Vis 16(5):429–457
Adobe Inc., Adobe Photoshop Elements, 4.0, http://www.adobe.co.uk/products/photoshopelwin/main.html
Google, Inc., Picasa Photo Management Software, 2.0. http://www.picasa.google.com/
Author information
Authors and Affiliations
Corresponding author
Appendix 1
Appendix 1
In this section, we give the specific results, per directory, for the large Corel-test set. Table 3 provides the results when considering four orientations and Table 4 provides the results when considering three orientations. Note that in each case, the top performing categories are largely outdoor photographs with very clear orientations. The worst performing categories are abstracts, textures, and backgrounds; these contain little or no orientation information. The ‘Army’ category which appears as the worst performer in both the three and four orientation case consisted of images that were four indoor machinery images, a single one which contained a person, and an additional single outdoor sunset photo.
When comparing Tables 3 and 4, as expected, many more categories are classified with higher accuracy in Table 4 (3 orientations). Only 17 of the 180 categories had the same performance between the 3 and 4 orientations (all of the others improved). Of the 17 categories, 4 did not improve because the performance was already 100%; the remaining 13 are listed below.
-
1.
Architecture_VIII (87/89)
-
2.
Contemporary_Buildings (82/90)
-
3.
British_Royalty (8/9)
-
4.
Architecture_III (58/66)
-
5.
Beaches (28/32)
-
6.
Children_II (5/6)
-
7.
Berlin (27/34)
-
8.
Dance (6/8)
-
9.
Couples_II (16/23)
-
10.
EMS_Rescue (2/4)
-
11.
Cowboys (7/14)
-
12.
Costumes (1/2)
-
13.
Army (1/5)
Rights and permissions
About this article
Cite this article
Baluja, S. Automated image-orientation detection: a scalable boosting approach. Pattern Anal Applic 10, 247–263 (2007). https://doi.org/10.1007/s10044-006-0059-1
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10044-006-0059-1