Gabor Transform-based Image Feature Extraction Method Study

Time-frequency analysis is one of the important means of signal processing. As the classic of Fourier transform time-frequency analysis tools, it establishes signal from the time to frequency domain. But it only reflects the whole frequency domain features of the signal in all the time, and can't provide any local time in frequency domain. Based on time-frequency analysis of Gabor transform, overcome the defects which traditional Fourier transform haven't any resolution in frequency domain, reflected signals the time-frequency joint. It is not only proved to be the optimal joint time-frequency resolution in the Heisenberg uncertainty principle, but also 2-D Gabor function can describe a mammal primary vision system of simple visual neurons, which has laid a theoretical basis for its application in image processing.The thesis based on Gabor time-frequency analysis, according to the image feature extraction, image registration, the image enhancement related problems research. The Gabor transformation has introduced in the foundation of preliminary understanding image feature extraction, Gabor transform of the principle and properties are discussed and applied. The experiments show that the algorithm with a traditional feature extraction algorithm has better performance and can extract image details and better visual inspection result. It has laid certain foundation for the later image registration and image enhancement processing.In image registration, by odd Gabor filters and even Gabor filters extracted image texture features and geometric features and combined with adaptive image registration, improved registration accuracy, especially in the details have been greatly improved. In the image enhancement, the algorithm is improved based on the original algorithm, Gabor is proposed based on the image enhancement algorithm, combined with image interpolation and the traditional image enhancement methods enhanced image, so that the images get better visual inspection result. Finally, image simulation test system image enhancement is verified.