Abstract:
There are certain changes in this world which fall below the visible spectrum of human beings, thus cannot be perceived with the naked eye. These subtle changes may contain useful information. Our goal is to magnify these subtle variations to a humanly acceptable video spectrum, to extract hidden information from them for further analysis, such as vital signs. Due to rapid development in technology, wide number of techniques have been proposed to magnify invisible changes, yet there is a need for efficient, cost effective and reliable video magnification technique. In this thesis, time efficient eulerian video magnification technique is proposed for different applications. The proposed technique utilizes the concept of time and spatial uniformity to reduce the computational complexity. Video may contain unwanted noise due to sensors used or camera jittering. The input frame of video sequence is passed through a wavelet denoising filter to remove/minimize the unwanted artifacts. It is then spatially decomposed and passed through a bandpass temporal filter to extract frequencies of our interest. The
filtered signal is magnified to enhance small motions and subtle variations in colour to reveal important information. Extracted magnified signals can be used in microscopic videos to amplify their invisible movements or to extract vital signs from the minute palpitations in wrist (radial artery), as heart pumps blood through body. Our proposed technique can also be used to observe rapid colorimetric changes in chemical reactions. Visual and quantitative comparison (with state of art existing technique) are performed to verify the signicance of proposed technique. Simulation results reveal that the proposed technique is almost two times efficient and more accurate as compared to state of art video magnification technique.
