Conventional relevance feedback (RF) schemes improve the performance of content-based image retrieval (CBIR) requiring the user to annotate a large number of images. To reduce the labeling effort of the user, this paper presents a novel active learning (AL) method to drive RF for retrieving remote sensing images from large archives in the framework of the support vector machine classifier. The proposed AL method is specifically designed for CBIR and defines an effective and as small as possible set of relevant and irrelevant images with regard to a general query image by jointly evaluating three criteria: uncertainty; diversity; and density of images in the archive. The uncertainty and diversity criteria aim at selecting the most informative images in the archive, whereas the density criterion goal is to choose the images that are representative of the underlying distribution of data in the archive. The proposed AL method assesses jointly the three criteria based on two successive steps. In the first step, the most uncertain (i.e., ambiguous) images are selected from the archive on the basis of the margin sampling strategy. In the second step, the images that are both diverse (i.e., distant) to each other and associated to the high-density regions of the image feature space in the archive are chosen from the most uncertain images. This step is achieved by a novel clustering-based strategy. The proposed AL method for driving the RF contributes to mitigate problems of unbalanced and biased set of relevant and irrelevant images. Experimental results show the effectiveness of the proposed AL method.

The above text is overhead of text which no engineer can understand easily.

Lets start analysis the above text.

The Content Based Remote Sensing Image Retrieval is the methos which is basically used in Satellite, Satellite sends geographical images to earth station. and maintaining the content of the given data in Database is huge headache in computer science which we call as content management system i.e Content-Based Remote Sensing Image Retrieval (CBRSIR).

Now Question is how images are optimized and quality of images good enough to ensure the quality test.

With the development of satellite technology, Large volume remote sensing images, become available. Conventional RS image retrieval systems often rely on keywords/tags in terms of sensor type, geographical location,and data acquisition time of images stored in the archives. The performance of tag matching-based retrieval approaches highly depends on the availability and the quality of manual tags.

Any CBIR system essentially consists of (at least) two modules
1) a feature extraction module that derives a set of features for characterizing and describing images
2) a retrieval module that searches and retrieves images similar to the query image.

Querying image contents from large RS data archives depends on the capability and effectiveness of the feature extraction techniques in describing and representing the images. In the RS literature, several primitive features have been presented for retrieval purposes, such as the following: intensity features, color features, shape features, texture features, and local invariant features.However, the low-level features from an image have a very limited capability in representing and analyzing the high-level concept conveyed by RS images (i.e., the semantic content of RS images). This issue is known as the semantic gap that occurred between the low-level features and the high-level semantic content and leads to poor CBIR performance.

An effective approach to reduce the annotation effort in RF is active learning (AL) that aims at finding the most informative images in the archive that, when annotated and included in the set of relevant and irrelevant images (i.e., the training set), can significantly improve the retrieval performance. Moreover, selecting the most informative images results in the following:

1) a smaller number of RF iterations to optimize the CBIR and
2) a reduced annotation time due to the optimization of the training set with a minimum number of highly informative images. In the RS community, most of the previous studies in AL have been developed in the context of classification problems for land-cover map generation.

AL has been marginally considered in the framework of CBIR problems in the RS community. To the best of our knowledge, only one AL method is presented, which is developed in the context of the support vector machine (SVM) classifier and inspired from AL methods used for classification problems.

To overcome the aforementioned critical issues, in this paper,we propose a CBIR approach that includes a novel triple criteria
AL (TCAL) method to drive RF in CBIR. For the selection of the most informative as well as representative unlabeled images of images to be annotated, the proposed TCAL method jointly evaluates three criteria:
1) uncertainty
2) diversity
3) density of images in the archive.

In order to assess the aforementioned three criteria, the proposed TCAL method exploits a two-step procedure defined in the framework of the SVM classifier. In the first step, the most uncertain (i.e.,ambiguous) images are selected by the well-known MS strategy, whereas in the second step, the diverse images among the most uncertain ones are selected from the highest density regions of the image feature space. The latter step is achieved by a novel clustering-based strategy that evaluates the density and diversity of unlabeled images in the image feature space to drive the selection of images to be annotated.

REFERENCES

1. A. Smeulders, M. Worring, S. Santini, A. Gupta, and R. Jain, “Content-based image retrieval at the end of the early years,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 22, no. 12, pp. 1349–1380, Dec. 2000.
2. R. Datta, D. Joshi, J. Li, and J.-Z. Wang, “Image retrieval: Ideas, influ- ences, trends of the new age,” ACM Comput. Surveys, vol. 40, no. 2,pp. 1–60, Apr. 2008.
3. P. Hong, Q. Tian, and T. S. Huang, “Incorporate support vector machines to content-based image retrieval with relevant feedback,” in Proc. IEEE Int. Conf. Image Process., Vancouver, BC, Canada, 2000, pp. 750–753.
4. X. S. Zhou and T. S. Huan, “Relevance feedback in image retrieval: A comprehensive review,” Multimedia Syst., vol. 8, no. 6, pp. 536–544, Apr. 2003.
5. Q. Bao and P. Guo, “Comparative studies on similarity measures for remote sensing image retrieval,” in Proc. IEEE Int. Conf. Syst., Man Cybern., Hague, The Netherlands, 2004, pp. 1112–1116.
6. T. Bretschneider, R. Cavet, and O. Kao, “Retrieval of remotely sensed imagery using spectral information content,” in Proc. IEEE Int. Geosci. Remote Sens. Symp., Toronto, ON, Canada, 2002, pp. 2253–2255.
7. T. Bretschneider and O. Kao, “A retrieval system for remotely sensed imagery,” in Proc. Int. Conf. Imag. Sci., Syst., Technol., Las Vegas, NV, USA, 2002, pp. 439–445.
8. G. Scott, M. Klaric, C. Davis, and C.-R. Shyu, “Entropy-balanced bitmap tree for shape-based object retrieval from large-scale satellite imagery databases,” IEEE Trans. Geosci. Remote Sens., vol. 49, no. 5, pp. 1603–1616, May 2011.
9. A. Ma and I. K. Sethi, “Local shape association based retrieval of infrared satellite images,” in Proc. IEEE Int. Symp. Multimedia, Irvine, CA, USA, 2005, pp. 551–557.
10. M. Ferecatu and N. Boujemaa, “Interactive remote-sensing image retrieval using active relevance feedback,” IEEE Trans. Geosci. Remote Sens., vol. 45, no. 4, pp. 818–826, Apr. 2007.
11. Y. Li and T. Bretschneider, “Semantics-based satellite image retrieval using low-level features,” in Proc. IEEE Int. Geosci. Remote Sens. Symp., Anchorage, AK, USA, 2004, vol. 7, pp. 4406–4409.
12. Y. Hongyu, L. Bicheng, and C. Wen, “Remote sensing imagery retrieval based-on Gabor texture feature classification,” in Proc. Int. Conf. Signal Process., Troia, Turkey, 2004, pp. 733–736.
13. B. S. Manjunath and W. Y. Ma, “Texture features for browsing and retrieval of image data,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 18, no. 8, pp. 837–842, Aug. 1996.
14. S. Newsam, L. Wang, S. Bhagavathy, and B. S. Manjunath, “Using texture to analyze and manage large collections of remote sensed image and video data,” J. Appl. Opt., vol. 43, no. 2, pp. 210–217, Jan. 2004.
15. A. Samal, S. Bhatia, P. Vadlamani, and D. Marx, “Searching satellite imagery with integrated measures,” Pattern Recognit., vol. 42, no. 11, pp. 2502–2513, Nov. 2009.
16. S. Newsam and C. Kamath, “Retrieval using texture features in high resolution multi-spectral satellite imagery,” in Proc. SPIE Defense Security Symp., Data Mining Knowl. Discov.—Theory, Tools, Technol. VI, Orlando,
    FL, USA, 2004, pp. 21–32.
17. Y. Yang and S. Newsam, “Geographic image retrieval using local invariant features,” IEEE Trans. Geosci. Remote Sens., vol. 51, no. 2, pp. 818–832, Feb. 2013.
18. L. Bruzzone, C. Persello, and B. Demir, “Active learning methods in classification of remote sensing images,” in Signal and Image Processing for Remote Sensing, 2nd ed, C. H. Chen, Ed. Boca Raton, FL, USA:CRC Press, 2012, ch. 15, pp. 303–323.
19. G. Schohn and D. Cohn, “Less is more: Active learning with support vector machines,” in Proc. 17th Int. Conf. Mach. Learn., Stanford, CA, USA, 2000, pp. 839–846.
20. S. Tong and D. Koller, “Support vector machine active learning with applications to text classification,” J. Mach. Learn. Res., vol. 2, pp. 45–66,2001.
21. K. Brinker, “Incorporating diversity in active learning with support vector machines,” in Proc. Int. Conf. Mach. Learn., Washington, DC, USA, 2003, pp. 59–66.
22. R. Zhang and A. I. Rudnicky, “A large scale clustering scheme for kernel k-means,” in Proc. IEEE Int. Conf. Pattern Recog., Quebec, QC, Canada, 2002, pp. 289–292.