dc.contributor.advisor | Havlicek, Joseph P | |
dc.creator | Nguyen, Chuong T. | |
dc.date.accessioned | 2019-04-27T21:32:10Z | |
dc.date.available | 2019-04-27T21:32:10Z | |
dc.date.issued | 2012 | |
dc.identifier | 99274850002042 | |
dc.identifier.uri | https://hdl.handle.net/11244/318933 | |
dc.description.abstract | The classical Fourier transform is the cornerstone of traditional linear | |
dc.description.abstract | signal and image processing. The discrete Fourier transform (DFT) and the | |
dc.description.abstract | fast Fourier transform (FFT) in particular led to | |
dc.description.abstract | profound changes during the later decades of the last century in how | |
dc.description.abstract | we analyze and process 1D and multi-dimensional signals. | |
dc.description.abstract | The Fourier transform represents a signal as an infinite superposition | |
dc.description.abstract | of stationary sinusoids each of which has constant amplitude and constant | |
dc.description.abstract | frequency. However, many important practical signals such as radar returns | |
dc.description.abstract | and seismic waves are inherently nonstationary. Hence, more complex | |
dc.description.abstract | techniques such as the windowed Fourier transform and the wavelet transform | |
dc.description.abstract | were invented to better capture nonstationary properties of these signals. | |
dc.description.abstract | In this dissertation, I studied an alternative nonstationary representation | |
dc.description.abstract | for images, the 2D AM-FM model. In contrast to the | |
dc.description.abstract | stationary nature of the classical Fourier representation, the AM-FM model | |
dc.description.abstract | represents an image as a finite sum of smoothly varying amplitudes | |
dc.description.abstract | and smoothly varying frequencies. The model has been applied successfully | |
dc.description.abstract | in image processing applications such as image segmentation, texture analysis, | |
dc.description.abstract | and target tracking. However, these applications are limited | |
dc.description.abstract | to \emph{analysis}, meaning that the computed AM and FM functions | |
dc.description.abstract | are used as features for signal processing tasks such as classification | |
dc.description.abstract | and recognition. For synthesis applications, few attempts have been made | |
dc.description.abstract | to synthesize the original image from the AM and FM components. Nevertheless, | |
dc.description.abstract | these attempts were unstable and the synthesized results contained artifacts. | |
dc.description.abstract | The main reason is that the perfect reconstruction AM-FM image model was | |
dc.description.abstract | either unavailable or unstable. Here, I constructed the first functional | |
dc.description.abstract | perfect reconstruction AM-FM image transform that paves the way for AM-FM | |
dc.description.abstract | image synthesis applications. The transform enables intuitive nonlinear | |
dc.description.abstract | image filter designs in the modulation domain. I showed that these filters | |
dc.description.abstract | provide important advantages relative to traditional linear translation invariant filters. | |
dc.description.abstract | This dissertation addresses image processing operations in the nonlinear | |
dc.description.abstract | nonstationary modulation domain. In the modulation domain, an image is modeled | |
dc.description.abstract | as a sum of nonstationary amplitude modulation (AM) functions and | |
dc.description.abstract | nonstationary frequency modulation (FM) functions. I developed | |
dc.description.abstract | a theoretical framework for high fidelity signal and image modeling in the | |
dc.description.abstract | modulation domain, constructed an invertible multi-dimensional AM-FM | |
dc.description.abstract | transform (xAMFM), and investigated practical signal processing applications | |
dc.description.abstract | of the transform. After developing the xAMFM, I investigated new image | |
dc.description.abstract | processing operations that apply directly to the transformed AM and FM | |
dc.description.abstract | functions in the modulation domain. In addition, I introduced two | |
dc.description.abstract | classes of modulation domain image filters. These filters produce | |
dc.description.abstract | perceptually motivated signal processing results that are difficult or | |
dc.description.abstract | impossible to obtain with traditional linear processing or spatial domain | |
dc.description.abstract | nonlinear approaches. Finally, I proposed three extensions of the AM-FM | |
dc.description.abstract | transform and applied them in image analysis applications. | |
dc.description.abstract | The main original contributions of this dissertation include the following. | |
dc.description.abstract | - I proposed a perfect reconstruction FM algorithm. I used a | |
dc.description.abstract | least-squares approach to recover the phase signal from its | |
dc.description.abstract | gradient. In order to allow perfect reconstruction of the phase function, I | |
dc.description.abstract | enforced an initial condition on the reconstructed phase. The perfect | |
dc.description.abstract | reconstruction FM algorithm plays a critical role in the | |
dc.description.abstract | overall AM-FM transform. | |
dc.description.abstract | - I constructed a perfect reconstruction multi-dimensional filterbank | |
dc.description.abstract | by modifying the classical steerable pyramid. This modified filterbank | |
dc.description.abstract | ensures a true multi-scale multi-orientation signal decomposition. Such a | |
dc.description.abstract | decomposition is required for a perceptually meaningful AM-FM image | |
dc.description.abstract | representation. | |
dc.description.abstract | - I rotated the partial Hilbert transform to alleviate rippling | |
dc.description.abstract | artifacts in the computed AM and FM functions. This adjustment results in | |
dc.description.abstract | artifact free filtering results in the modulation domain. | |
dc.description.abstract | - I proposed the modulation domain image filtering framework. I | |
dc.description.abstract | constructed two classes of modulation domain filters. I showed that the | |
dc.description.abstract | modulation domain filters outperform traditional linear shift | |
dc.description.abstract | invariant (LSI) filters qualitatively and quantitatively in applications | |
dc.description.abstract | such as selective orientation filtering, selective frequency filtering, | |
dc.description.abstract | and fundamental geometric image transformations. | |
dc.description.abstract | - I provided extensions of the AM-FM transform for image decomposition | |
dc.description.abstract | problems. I illustrated that the AM-FM approach can successfully | |
dc.description.abstract | decompose an image into coherent components such as texture | |
dc.description.abstract | and structural components. | |
dc.description.abstract | - I investigated the relationship between the two prominent | |
dc.description.abstract | AM-FM computational models, namely the partial Hilbert transform | |
dc.description.abstract | approach (pHT) and the monogenic signal. The established relationship | |
dc.description.abstract | helps unify these two AM-FM algorithms. | |
dc.description.abstract | This dissertation lays a theoretical foundation for future nonlinear | |
dc.description.abstract | modulation domain image processing applications. For the first time, one | |
dc.description.abstract | can apply modulation domain filters to images to obtain predictable | |
dc.description.abstract | results. The design of modulation domain filters is intuitive and simple, | |
dc.description.abstract | yet these filters produce superior results compared to those of pixel | |
dc.description.abstract | domain LSI filters. Moreover, this dissertation opens up other research problems. | |
dc.description.abstract | For instance, classical image applications such as image segmentation and | |
dc.description.abstract | edge detection can be re-formulated in the modulation domain setting. | |
dc.description.abstract | Modulation domain based perceptual image and video quality assessment and | |
dc.description.abstract | image compression are important future application areas for the fundamental | |
dc.description.abstract | representation results developed in this dissertation. | |
dc.format.extent | 218 pages | |
dc.format.medium | application.pdf | |
dc.language | en_US | |
dc.relation.requires | Adobe Acrobat Reader | |
dc.subject | Image processing | |
dc.subject | Radio frequency modulation | |
dc.subject | Amplitude modulation | |
dc.title | Modulation Domain Image Processing | |
dc.type | text | |
dc.type | document | |
dc.thesis.degree | Ph.D. | |
ou.group | College of Engineering::School of Electrical and Computer Engineering | |