Multi-Scale AM-FM Demodulation for Image Reconstruction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current image and video processing techniques, such as STFT and DWT, face challenges in effectively demodulating amplitude-modulation frequency-modulation (AM-FM) components for stationary and non-stationary signals, leading to inefficiencies in applications like fingerprint identification, retinal disease diagnosis, and cardiac image segmentation.
Innovation Solution
The implementation of a multi-scale AM-FM demodulation system using a multi-scale filterbank, instantaneous frequency estimation based on variable-spacing local linear phase, and multi-scale least square reconstruction, which computes extended analytic signals and estimates instantaneous amplitude, phase, and frequency, improving accuracy and reducing interference.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If STFT is used for non-stationary signals, then frequency content information at different time intervals is obtained, but the method cannot be effectively generalized to images and videos producing high-dimensional representations
Solution Approach 1:
The patent applies segmentation by dividing the frequency spectrum into multiple scales using a filterbank approach. Instead of computing a full high-dimensional STFT representation, the method segments frequencies into distinct bands (low, medium, high) and processes each band separately through scale-specific filters, thereby obtaining frequency content information without generating prohibitively high-dimensional representations for images and videos.
Solution Approach 2:
The patent transforms the problem from time-frequency analysis (STFT) to scale-frequency analysis. By introducing a multi-scale decomposition framework where signals are analyzed at different frequency scales rather than time intervals, the method avoids the dimensionality explosion of STFT while still capturing non-stationary characteristics through scale-specific instantaneous frequency estimates.
2Measurement precision
If DWT is used for image processing, then space-frequency representation is achieved with logarithmic frequency division, but frequency content is not measured directly
Solution Approach 1:
The patent substitutes the mechanical wavelet decomposition process with a signal processing approach based on analytic signals and instantaneous frequency estimation. Instead of relying on wavelet basis functions to indirectly represent frequency content, the method uses Hilbert transforms and phase derivatives to directly compute instantaneous frequency, providing explicit frequency measurements while maintaining multi-scale analysis capabilities.
Solution Approach 2:
The patent changes the fundamental parameter from wavelet scale (logarithmic division) to instantaneous frequency (linear frequency measurement). By transforming the signal into analytic form and computing the derivative of the phase, the method directly measures frequency content in Hertz rather than wavelet scales, making frequency interpretation more intuitive and operationally easier.
3Measurement precision
If multi-scale filterbank is used for AM-FM demodulation, then accurate instantaneous frequency estimates are obtained, but computational complexity increases
Solution Approach 1:
The patent reduces computational complexity through segmentation by dividing the frequency spectrum into distinct bands using a multi-scale filterbank. Each band is processed independently with scale-appropriate filters, allowing accurate instantaneous frequency estimation within each band without requiring computationally intensive full-spectrum analysis. This segmented approach maintains precision while reducing overall computational burden.
Solution Approach 2:
The patent applies partial action by focusing computational resources on estimating instantaneous frequency within specific frequency bands rather than attempting to analyze the entire spectrum uniformly. The multi-scale filterbank selectively processes only the relevant frequency components for each scale, avoiding excessive computation on frequencies outside the current band of interest while still achieving accurate demodulation.
Data Source
AI summary
Image and video processing using multi-scale amplitude-modulation frequency-modulation (“AM-FM”) demodulation where a multi-scale filterbank with bandpass filters that correspond to each scale are used to calculate estimates for instantaneous amplitude, instantaneous phase, and instantaneous frequency. The image and video are reconstructed using the instantaneous amplitude and instantaneous frequency estimates and variable-spacing local linear phase and multi-scale least square reconstruction techniques. AM-FM demodulation is applicable in imaging modalities such as electron microscopy, spectral and hyperspectral devices, ultrasound, magnetic resonance imaging (“MRI”), positron emission tomography (“PET”), histology, color and monochrome images, molecular imaging, radiographs (“X-rays”), computer tomography (“CT”), and others. Specific applications include fingerprint identification, detection and diagnosis of retinal disease, malignant cancer tumors, cardiac image segmentation, atherosclerosis characterization, brain function, histopathology specimen classification, characterization of anatomical structure such as carotid artery walls and plaques or cardiac motion and as the basis for computer-aided diagnosis to name a few.