Abstract

In this work, we propose an efficient algorithm,which utilizes the combination of discrete cosine transform (DCT)and phase correlation (PC) for fast object detection. To test thealgorithm´s classification performance and computationalcomplexity we developed a prototype and conducted severalexperiments with a public available car dataset. Furthermore, wecompared our experimental results to a state-of-the-art objectdetection method. The proposed method uses the energycompaction property of DCT and requires less number ofcoefficients than fast Fourier transformation (FFT)-basedtechniques to compute PC. The computational complexity andmemory requirements are significantly reduced using this method.According to our results, the proposed algorithm outperforms thebaseline method with respect to training time and classificationaccuracy.

Reference

Eidenberger, H., & Hörhan, M. (2016). An Efficient DCT template-based Object Detection Method using Phase Correlation. In Asilomar Conference on Signals, Systems, and Computers Proceedings MA8b3-4. Asilomar Conference on Signals, Systems, and Computers, Pacific Grove, Non-EU. http://hdl.handle.net/20.500.12708/56633