The Field-Of-View(FOV) of compact cameras are very less compared to the FOV of humans. A single shot of capturing compact camera sensor does not capture the scenes as the humans vision. Mosaicing is a procedure of uniting the partial parts of a scene to form a large image that represents the whole scene. The main aim of mosaicing is to increase the FOV and to trace the path of any moving objects. In this paper, a real-time mosaicing is implemented with the various algorithms and compared with the proposed algorithm. The proposed method is a combination of FAST detector and Steerable filters. All the algorithms are executed with the multiple rapid prototype hardware’s and compared their performance interms of many parameters like corner detection rate, image mosaicing time, video mosaicing execution time. Geekbench mark test of having various parameters is conducted on three hardware’s. Our experimental results proved that the intel compute stick and Lenovo ideacentre 300 hardwares have high mobility to capture the remote areas of a scene or low altitude images to construct the mosaic image.
There are some situations where the scene is beyond the scope of the camera. Compact cameras has low FOV compare to the human visual system because the Field-Of-View(FOV) of compact cameras are around the 35×55 degrees where as human being has 135×180 degrees of FOV. In this scenarios mosaic procedure is applied to construct the scene with more FOV and the image obtain the by this process is called the mosaic image. Image mosaic is a procedure of uniting partial images of a scene to produce large mosaic image. Feature detectors detects all the keypoints of an image and descriptors extract the data around each corner point. Feature based mo-saic algorithms have some common phases they are capturing the real-time scene, interest points detections, extracting the information around the corner points, comparing the feature matching using the hamming distance, mapping the matched corner points by using geometric measurement and warping and blending. Mosaicing is constructed with the SUSAN, Harris, KLT, FAST
and proposed methods. All algorithms are compared interms of accuracy, recall and true cornerness of synthetic image. Farzin M. and Farahnaz M. computed the performance of various algorithms interms of accuracy. So many authors introduced many algorithms each have their own advantage and disadvantages. Harris, SUSAN, KLT, Tomsai and Shi and FAST algorithms are familiar feature detectors. Among above menctioned detectors, only FAST algorithm is best suited for real-time applications like Simultanious-Localization And Mapping. Harris detector is an efficient method but it has more computation time compare to the FAST method. SUSAN if purely applicable for non-real-time scenarios because it takes much time for corner detection than any other algorithms. In this paper a real-time image mosaicing and video mosaicing is implemented with the existing methods and as well as with the proposed method with three hardwares. The proposed algorithm is a combination of FAST detector and steerable filters. The FAST detector detects the feature points and steerable filters are introduced to steer the image in any orientation. Steerable filters main advantage is selective orientation.
A real-time mosaicing of each algorithm is executed with the three hardware devices and their performance is compared interms of corner detection rate, image mosaicing computation time and video mosaicing time. The hardware performance is measures with the GeekBench mark tool. Though the perfor-mance of laptop is high but the mobility (Dimensions and weight) is poor when compared with the intel compute stick and Lenovo ideacentre 300 on board computer devices. this sticks are used to capture the remote images and construct the mosaic image in real-time. This compute sticks can be stick to any moving vehicle like drone and capture the remote images. so need of specialized hardwares required for computer vision applications.
In this paper, an innovative method was implemented with the general-purpose hardware devices. The main advantage of this approach is no need of specialized hardwares for the mosaicing of remote images. The proposed method was 1.44 times faster than Harris corner detector, 5.24 times faster than KLT method and 10.4 times quicker than SUSAN technique. Corner detection rate is calculated with all the detectors and hardwares. Frames of a video are used to mosaic and obtain mosaic image with the three hardware devices by the FAST and proposed methods. Geek bench test is conducted for the three hardware devices to calculate the performance. Experi-mental results proved that proposed method is best suited for the real-time image mosaicing with an efficient accuracy, recall and true cornerness parameters and with extended field-of-view.
- Mahesh, Subramanyam M.V. Automatic feature based image registration using SIFT algorithm. ICCNT-12; Coimbatore. 2012 Jul. p. 26-8.
- S.V.P., Jilani, Ramana R.P., A Critical Review on Image mosaic-ing.International Conference on Computer Communication and Informat-ics , Jan.07-09, 2016, Coimbatore, INDIA.
- Farzin M. and F. Mohanna, Performance evaluation of corner detectors using consistency and accuracy measures Computer Vision and Image Understanding Volume-102, pages 81-94, 2006.
- Rosten and T. Drummond , Machine learning for high speed corner detection. in 9th European Conference on Computer Vision, vol. 1, 2006, pp.430-443.
- David A. Forsyth, Jean Ponce., Computer vision: A Modern Approach. Prentice Hall, August 2002.
- Mallick and S. Prakash, Feature based image mosaicking, Department of Electrical and Computer Engineering, University of California, San Diego, 2002.
- S.V. Prathap, Jilani S.A.K and P.R.Reddy A Real-Time Image Mosaicing Using FAST Detector and FREAK Descriptor IEEE WISPNET 2017 conference.PP.2444-2449.DOI.978-1-5090-4442-9/17/$31.00 c 2017 IEEE
- M. Smith and J. M. Brady (May 1997). SUSAN-a new approach to low level image processing. International Journal of Computer Vision. 23 45-78.DOI:10.1023/A:1007963824710.
- Carlo Tomasi and Takeo Kanade. Detection and Tracking of Point Features. Carnegie Mellon University Technical Report CMU-CS-91-132, April 1991.
- 10. H. Moravec, Obstacle avoidance and navigation in the real world by a seeing robot rover, Tech Report CMURI-TR3 Carnegie-Mellon University, Robotics Institute, 1980.
- Harris and Stephens, A combined corner and edge detector, in Proceedings of the 4th Alvey Vision Conference, pp. 147-151, 1988.
- Nigar and Mahesh, Image mosaic using fast corner detection, Intern. Journal of Advanced Research in Electronics and Communication Engineering, vol. 1, no. 6, 2012.
- Mahesh, M.V. Subramanyam. Feature Based Image Mosaic Using Steerable Filters and Harris Corner Detector I.J. Image, Graphics and Signal Processing, 2013, 6, 9-15 DOI: 10.5815/ijigsp.2013.06.02
- Mahesh, M.V. Subramanyam. Invariant Corner Detection Using Steerable Filters and Harris Algorithm. Signal & Image Processing: An International Journal (SIPIJ) Vol.3, No.5, October 2012.
- J. Davis, Mosaics of scenes with moving objects, in Proc. CVPR, 1998.
- R. Szeliski and H. Y. Shum, Creating full view panoramic image mosaic and environment maps, in Proc. SIGGRAPH’97, 1997.
- E. Bostanci, Is hamming distance only way for matching binary image feature descriptors? Electronics Letters, vol. 50, no. 11, pp. 806-808, May 2014.
- Assaf Zomet, Anat Levin, Shmuel Peleg, and Yair Weiss, Seamless image stitching by minimizing false edges, IEEE Transactions on Image Processing, vol. 15, no. 4, Apr. 2006.
- M. A. Fischler and R. C. Bolles, Random sample consensus: A paradigm for model ﬁtting with applications to image analysis and automated cartography, C. of the ACM, Ed. Communications of the ACM, June 1981, vol. 24, no. 6.
384 total views, 1 views today