Chuong T. Nguyen

Modulation Domain Template Tracking

For the first time, we perform normalized correlation template tracking in the modulation domain. For each frame of the video sequence, we compute a multi-component AM-FM image model that characterizes the local texture structure of objects and backgrounds. Tracking is carried out by formulating a modulation domain correlation function in the derived feature space. Using visible and longwave infrared sequences as illustrative examples, we study the performance of this new approach relative to two basic pixel domain correlation template trackers. We also present preliminary results from a new dual domain tracker that operates simultaneously in both the pixel and modulation domains.

AM-FM image filters

We introduce a multicomponent invertible AM-FM image transform and use it to define new nonlinear AM-FM filters for performing modulation domain image processing. The key elements of the transform are analysis and synthesis filterbanks based on the steerable image pyramid and perfect reconstruction demodulation algorithms based on analytic differentiation of continuous cubic tensor spline models fit to the unwrapped phase samples of a digital image. We demonstrate spatially and spectrally localized orientation and frequency selective filtering, simple image restoration, and image fusion in the modulation domain. These results are also among the first to demonstrate high fidelity image reconstructions from computed multicomponent AM-FM models.

Infrared Target Tracking with AM-FM Consistency Checks

Challenging infrared data sequences such as the well-known AMCOM closure sequences are characterized by highly nonstationary evolutionary target and clutter signatures, poor target-to- clutter ratios, and complex kinematics arising from both the target motion and the motion of the sensor platform itself. In such cases, track consistency checks can provide a valuable means for detecting an imminent track loss. In this paper, we consider a simple target model with a correlation-based detection process and a straightforward SIR particle filter track processor. We show that the performance of the track processor can be dramatically improved by incorporating modulation domain consistency checks to identify failure in the correlation-based detection process. This strategy results in a robust dual-domain tracker that, despite the simplicity of its state model, delivers superior tracking performance against the very difficult AMCOM sequences.

Modulation Domain Features for Discriminating Infrared Targets and Backgrounds

For the first time, we compute modulation domain features for infrared targets and backgrounds, including dominant modulations that characterize the local texture contrast, orientation, and granularity. We present a practical computational approach and introduce a new FM algorithm designed to reduce the approximation errors characteristic of many existing discrete techniques. By performing experiments against actual FLIR approach sequences, we verify that typical IR imagery does indeed possess sufficient texture structure for effective modulation domain characterization. We demonstrate qualitatively that the modulation domain features can significantly enhance target-background class separability relative to pixel domain features.

Online Consistency Checking for AM-FM Target Tracks

We compute AM-FM models for infrared video frames depicting military targets immersed in structured clutter backgrounds. We show that independent correlation based detection processes can be implemented in the pixel and modulation domains and used to construct useful online track consistency checks that indicate when the detection process has been degraded due to nonstationary evolution of the target signature. Throughout the paper, we use the well-known AMCOM closure sequences as exemplars.

FM filters for modulation domain image processing

For the first time, we demonstrate modulation domain image filters that achieve perceptually motivated image processing goals by directly manipulating the FM functions in a multi-component AM-FM image model. The action of previous modulation domain filters has been limited to modification of the AM functions based on the values of the AM and FM functions. This is because reconstruction of the modified phase from the filtered frequency modulation vectors was an unsolved problem. Here, we present two new algorithms capable of reconstructing the phase from the processed frequencies, one based on a least squares solution of the discrete Poisson equation with Neumann boundary condition and one based on cubic tensor product spline integration. New modulation domain FM filters are designed to modify both the orientations and magnitudes of the visually important emergent image frequency vectors. In our most dramatic example, we demonstrate an FM filter that autonomously changes the stripes on the pants in the well known Barbara image from vertical to horizontal.

Modulation domain infrared target models

We compute joint AM-FM models that characterize infrared targets and backgrounds in the modulation domain. We consider spatially localized structures within an IR image as sums of nonstationary, quasi-sinusoidal functions admitting locally narrowband amplitude and frequency modulations. By quantitatively estimating the modulations that dominate the signal spectrum on a spatially local basis, we obtain a new modulation domain feature vector that can augment the more traditional pixel domain, Fourier spectrum, and multispectral color features that have been used in IR target detection and tracking systems for a long time. Our preliminary studies, based primarily on midwave and longwave missile approach sequences, suggest that IR targets and backgrounds do typically possess sufficient spatially local modulated structure (i.e., texture) for modulation domain techniques to be meaningfully applied. We also present qualitative results strongly indicating that the modulation domain feature vector is a powerful tool for discriminating infrared targets and backgrounds.

First Results in Perceptually-Based AM-FM Image Filtering

We combine an adaptation of the steerable image pyramid sub- band decomposition with a spline-based perfect reconstruction demodulation algorithm to obtain an invertible AM-FM image transform. For the first time, we achieve perceptually-based signal processing goals by applying filtering operations directly to the computed subband amplitude and frequency modulations. The results are dramatic and would be difficult or impossible to obtain by linear processing. In our most interesting example, a simple AM-FM filter succeeds in smoothly and naturally removing the bands from the hat in the well-known Lena image.

AM-FM models, partial Hilbert transform, and the monogenic signal

We propose a new algorithm to compute the frequency modulation functions associated with the well-known monogenic signal. The new algorithm extracts the frequency modulation functions without requiring an auxiliary process to estimate local orientation. In addition, we show that, in situations where a multi-scale multi-orientation decomposition is required to analyze a signal, the partial Hilbert transform approach computes AM-FM functions similar to those obtained by the monogenic signal while maintaining a more efficient signal representation.

FM processing with generalized amplitude & phase: Application to modulation domain geometric image transformations

We introduce new generalized AM and FM functions to perform nonlinear image filtering in the modulation domain with consistent, artifact free phase reconstruction. The new framework enables us to design nonlinear filters in the modulation domain that are capable of producing perceptually motivated signal processing results. As an illustration, we demonstrate that the modulation domain geometric image transformations designed under this framework deliver artifact-free results that are consistent with those of classical intensity-based geometric image transformations.