| Summary: | 碩士 === 國立雲林科技大學 === 機械工程系碩士班 === 101 === In recently, due to camera trend to miniaturization, images have more disorderly motion components by incidentally shake of body when hand-held video devices captures image sequences. To overcome this problem, this thesis use digital image stabilization method to solve the effect in troublesome unstable image sequences.
Digital image stabilization is composed of three major subjects: Motion Estimation, Motion Smoothing and Image Compensation. Motion estimation is that sensor has acquired image information to execute the matching of every pair frames to get hand motion signal of photographer. With fast search mode of estimating search method, it can reduce calculating amount of system, and can output motion signal in real-time that will not make system delay; Motion smoothing is using digital signal processing to effectively divide hand motion into High frequency Shaking Noise and Low frequency Autonomy Motion; Image compensation is that reverse compensates high frequency shaking noise then keeps the low frequency autonomy motion. Finally, it output a stable image sequences.
With the progress of science and technology, shooting resolution of video devices is higher and higher. Based on identical shake, motion estimation of higher shooting resolution is what caused huge calculating among, therefore this thesis based on block matching that purposes a novel motion estimation search algorithm to enhance the operation efficiency of motion estimation when has large displacement. In motion smoothing, using Fourier Linear Combiner fit the low frequency autonomy motion signal and then using the motion signal of actual estimation subtract that get the high frequency shaking noise. If it wanted image output has best stable efficiency, it must finds the optimal parameter of motion smoothing ahead. Another object of this thesis is finding optimal parameter of Fourier Linear Combiner. Finally, it appeared a fast and stable image stabilization system.
|
|---|
