Abstract:We introduce a Prony-like method to recover a continuous domain 2-D piecewise smooth image from few of its Fourier samples. Assuming the discontinuity set of the image is localized to the zero level-set of a trigonometric polynomial, we show the Fourier transform coefficients of partial derivatives of the signal satisfy an annihilation relation. We present necessary and sufficient conditions for unique recovery of piecewise constant images using the above annihilation relation. We pose the recovery of the Fourier coefficients of the signal from the measurements as a convex matrix completion algorithm, which relies on the lifting of the Fourier data to a structured low-rank matrix; this approach jointly estimates the signal and the annihilating filter. Finally, we demonstrate our algorithm on the recovery of MRI phantoms from few low-resolution Fourier samples.
What problem does this paper attempt to address?
This paper attempts to address the problem of recovering 2D piecewise smooth images from a small number of Fourier samples. Specifically, the authors propose a Prony-like method to recover 2D piecewise smooth images in the continuous domain by leveraging the assumption that the set of image discontinuities (i.e., edges) can be confined to the zero level set of some trigonometric polynomial.
### Main Contributions:
1. **Theoretical Contributions**:
- The authors prove that the Fourier transform coefficients of the partial derivatives of piecewise smooth images satisfy an annihilation relation.
- They provide the necessary and sufficient conditions for the unique recovery of piecewise constant images.
- A method based on a convex matrix completion algorithm is proposed, which lifts the Fourier data to a structured low-rank matrix, thereby jointly estimating the signal and the annihilating filter.
2. **Algorithmic Contributions**:
- A single-step convex optimization algorithm is proposed for recovering MRI images from a small number of low-resolution Fourier samples.
- The algorithm is capable of jointly recovering the image and the annihilating polynomial without the need for explicit estimation of the annihilating polynomial.
### Application Background:
- **Medical Imaging**: Particularly in Magnetic Resonance Imaging (MRI), recovering high-quality images from a small number of low-resolution Fourier samples has significant practical importance, as it can reduce scan time and improve imaging efficiency.
### Method Overview:
1. **Signal Model**:
- Consider a class of 2D piecewise smooth functions where the function is smooth within each region and discontinuous at the boundaries.
- Assume that the set of discontinuities (edges) can be represented by the zero level set of a trigonometric polynomial.
2. **Annihilation Property**:
- It is proven that the Fourier transform coefficients of the partial derivatives of piecewise smooth signals satisfy an annihilation relation, meaning these coefficients can be annihilated by convolution with the Fourier coefficients of a trigonometric polynomial.
- This property allows the recovery of the edge set of the image from a finite number of Fourier samples.
3. **Recovery Algorithm**:
- The recovery problem is transformed into a structured low-rank matrix completion problem.
- A convex optimization method is used to solve it by minimizing the rank of the matrix to recover the signal.
- To handle noise and model mismatch, a convex relaxation method is proposed, solving it through nuclear norm minimization and projection operators.
### Experimental Results:
- The authors conducted experiments on simulated data, demonstrating that the algorithm can accurately recover MRI images from a small number of low-resolution Fourier samples.
- The experimental results show that the method can effectively recover images even with 20-fold undersampling, although some ringing artifacts are present, which can be removed with slight post-processing.
### Conclusion:
- This paper proposes a novel method for recovering 2D piecewise smooth images from a small number of low-frequency Fourier samples, particularly deriving the necessary and sufficient conditions for recovering the edge set for piecewise constant signals.
- The proposed single-step algorithm performs excellently in practical applications, especially in MRI imaging, significantly reducing scan time and improving imaging efficiency.