An Algorithm to Estimate Scale Weights of Complex Wavelets for Effective Feature Extraction in Aerial Images

Shubha Bhat; Ramesh D.R. Babu; Krishnan Rangarajan; Ramakrishna K.A.

doi:10.14429/dsj.64.7785

Shubha Bhat Dayananda Sagar College of Engineering, Bengaluru
Ramesh D.R. Babu Dayananda Sagar College of Engineering, Bengaluru
Krishnan Rangarajan Dayananda Sagar College of Engineering, Bengaluru
Ramakrishna K.A. Dayananda Sagar College of Engineering, Bengaluru

DOI: https://doi.org/10.14429/dsj.64.7785

Keywords: Dual-tree complex wavelet transform, feature detection, scale weights, precision-factor, recall-factor, F-measure

Abstract

Research on vision-based navigation of unmanned air vehicles (UAVs) has been in focus in the recent years, both in military and civilian application domains. Vision-based navigation uses features as cues to match successive images and compute the location of the UAV. The method proposed in this paper is an extension to the existing work to automatically compute scale weights of dual-tree complex wavelet transform (DTCWT) coefficients to improve the performance in terms of detection of features as against discrete wavelets or its Fourier counterparts. Existing DTCWT coefficients give improved time-shifted sensitivity, better directional selectivity with local phase information and limited redundancy. However, the drawback of this method is that the results depend on appropriate selection of scale weights α and β, which are different for different scenes. The existing technique incorporates a rule of thumb-based recommended values of α and β, irrespective of the scene which generates too many key-points, making the registration process computationally intensive. As a real-time application, the need of the problem is to extract less number of strong features. Thus, an algorithmic approach to compute optimal values of scale weights considering precision and accuracy is proposed. The method is tested on various synthetic, simulated, and aerial images having different transformations and in the presence of noise between successive images. It is observed that DTCWT descriptor performs the best and gives better results than scale invariant feature transform and Speeded up robust features.

Science Journal, Vol. 64, No. 6, November 2014, pp.549-556, DOI:http://dx.doi.org/10.14429/dsj.64.7785

Author Biographies

Shubha Bhat, Dayananda Sagar College of Engineering, Bengaluru

Ms Shubha Bhat received her B.E.(Elect. and Electronics Engg.) and M.S.(Comp. Sci. Engg.) in from Mangalore University, India and Technical University of Eindhoven, Netherlands, in 2000 and 2006, respectively. She is currently pursuing her PhD in Computer Vision and Image Processing from Visvesvaraya Technological University, Belgaum, India. She is currently working as an assistant professor at Dayananda Sagar college of Engineering, Bangalore. Her current research interests include : Computer vision, image processing and pattern recognition.

Ramesh D.R. Babu, Dayananda Sagar College of Engineering, Bengaluru

Dr Ramesh Babu D.R. received his B.E (Instrumentation Technology) and M.tech. (Industrial Technology) from University of Mysore, India in 1996 and 1999, respectively. He received his PhD in computer science – image processing from University of Mysore, India in 2004. He is currently the Head and Professor of Computer Science and Engineering Department at Dayananda Sagar College of Engineering, Bangalore, India. His current research interests include : Medical image processing, computer vision and pattern recognition.

Krishnan Rangarajan, Dayananda Sagar College of Engineering, Bengaluru

Dr Krishnan Rangarajan received his BE (hons) (Mechanical engineering) and MTech (Comp. Sci. Engg.) from REC Tiruchirapalli, India and IIT Delhi, India in 1983 and 1985, respectively. He received his PhD in computer science from University of Central Florida, Orlando, USA in 1990. He is currently a Professor in Computer Science and Engineering Department at Dayananda Sagar College of Engineering, Bangalore, India. His current research interests include : Object tracking in the field of computer vision and pattern recognition.

Ramakrishna K.A., Dayananda Sagar College of Engineering, Bengaluru

Prof. K.A. Ramakrishna received his MS (Comp. Sci. Engg.) from IIT Madras, India and MSc (Mathematics) from Bangalore University, India, in 1984 and 1970, respectively. He worked as Scientist ‘G’ Group director at Aeronautical development Establishment, Bangalore. He is currently a Professor in Computer Science and Engineering Department at Dayananda Sagar College of Engineering, Bangalore, India. His current research interests include : Computer vision and pattern recognition.