Measurement-Adaptive Sparse Image Sampling and Recovery (original) (raw)

Adaptive Sparse Image Sampling and Recovery

IEEE Transactions on Computational Imaging, 2018

This paper presents an adaptive and intelligent sparse model for digital image sampling and recovery. In the proposed sampler, we adaptively determine the number of required samples for retrieving image based on space-frequency-gradient information content of image patches. By leveraging texture in space, sparsity locations in DCT domain, and directional decomposition of gradients, the sampler structure consists of a combination of uniform, random, and nonuniform sampling strategies. For reconstruction, we model the recovery problem as a two-state cellular automaton to iteratively restore image with scalable windows from generation to generation. We demonstrate the recovery algorithm quickly converges after a few generations for an image with arbitrary degree of texture. For a given number of measurements, extensive experiments on standard image-sets, infra-red, and mega-pixel range imaging devices show that the proposed measurement matrix considerably increases the overall recovery performance, or equivalently decreases the number of sampled pixels for a specific recovery quality compared to random sampling matrix and Gaussian linear combinations employed by the state-of-the-art compressive sensing methods. In practice, the proposed measurement-adaptive sampling/recovery framework includes various applications from intelligent compressive imaging-based acquisition devices to computer vision and graphics, and image processing technology. Simulation codes are available online for reproduction purposes.

Effective Image Reconstruction Using Various Compressed Sensing Techniques

IEEE Conference , 2024

The prevailing information relies heavily on the principles articulated by Shannon and Nyquist. These theories assert that, to faithfully reconstruct a signal without distortion, the sampling frequency must exceed twice the signal's maximum frequency to prevent aliasing. Following this, signals undergo compression to eliminate inherent redundancies before transmission over the channel. Despite the effectiveness of this approach, a significant drawback arises from the substantial processing overhead involved in sampling and compression. This limitation renders the scheme unsuitable for contemporary applications, given the constraints of current computational capabilities. Compressive Sensing (CS) introduces a novel framework, grounded in signal decomposition and approximation theory. Serving as an alternative to the Nyquist criteria, CS presents advantages such as reduced sensing time and sampling rate. In this approach, signals are sampled below the Nyquist rate through linear projection onto a random basis, ensuring exact reconstruction of the original signal. CS contributes to the reduction of power consumption and computational complexity in handling digital data. The extraction of information is facilitated by the utilization of a sensing matrix. In the context of image restoration, the CS framework provides an innovative approach to recover highquality images from their compressed measurements. The success of image restoration in compressive sensing relies on the development of robust algorithms that can efficiently exploit sparsity while addressing the inherent trade-off between accuracy and computational complexity. We survey state-of-theart reconstruction algorithms, including iterative methods, convex optimization, and deep learning approaches, showcasing their strengths and limitations in restoring images from compressed measurements. The success of image restoration in compressive sensing relies on the development of robust algorithms that can efficiently exploit sparsity while addressing the inherent trade-off between accuracy and computational complexity. We survey state-of-the-art reconstruction algorithms, including iterative methods, convex optimization, and deep learning approaches, showcasing their strengths and limitations in restoring images from compressed measurements.

Improved Compressive Sensing of Natural Scenes Using Localized Random Sampling

Scientific Reports, 2016

Compressive sensing (CS) theory demonstrates that by using uniformly-random sampling, rather than uniformly-spaced sampling, higher quality image reconstructions are often achievable. Considering that the structure of sampling protocols has such a profound impact on the quality of image reconstructions, we formulate a new sampling scheme motivated by physiological receptive field structure, localized random sampling, which yields significantly improved CS image reconstructions. For each set of localized image measurements, our sampling method first randomly selects an image pixel and then measures its nearby pixels with probability depending on their distance from the initially selected pixel. We compare the uniformly-random and localized random sampling methods over a large space of sampling parameters, and show that, for the optimal parameter choices, higher quality image reconstructions can be consistently obtained by using localized random sampling. In addition, we argue that the localized random CS optimal parameter choice is stable with respect to diverse natural images, and scales with the number of samples used for reconstruction. We expect that the localized random sampling protocol helps to explain the evolutionarily advantageous nature of receptive field structure in visual systems and suggests several future research areas in CS theory and its application to brain imaging. Sampling protocols have drastically changed with the discovery of compressive sensing (CS) data acquisition and signal recovery 1,2. Prior to the development of CS theory, the Shannon-Nyquist theorem determined the majority of sampling procedures for both audio signals and images, dictating the minimum rate, the Nyquist rate, with which a signal must be uniformly sampled to guarantee successful reconstruction 3. Since the theorem specifically addresses minimal sampling rates corresponding to uniformly-spaced measurements, signals were typically sampled at equally-spaced intervals in space or time before the discovery of CS. However, using CS-type data acquisition, it is possible to reconstruct a broad class of sparse signals, containing a small number of dominant components in some domain, by employing a sub-Nyquist sampling rate 2. Instead of applying uniformly-spaced signal measurements, CS theory demonstrates that several types of uniformly-random sampling protocols will yield successful reconstructions with high probability 4-6. While CS signal recovery is relatively accurate for sufficiently high sampling rates, we demonstrate that, for the recovery of natural scenes, reconstruction quality can be further improved via localized random sampling. In this new protocol, each signal sample consists of a randomly centered local cluster of measurements, in which the probability of measuring a given pixel decreases with its distance from the cluster center. We show that the localized random sampling protocol consistently produces more accurate CS reconstructions of natural scenes than the uniformly-random sampling procedure using the same number of samples. For images containing a relatively large spread of dominant frequency components, the improvement is most pronounced, with localized random sampling yielding a higher fidelity representation of both low and moderate frequency components containing the majority of image information. Moreover, the reconstruction improvements garnered by localized random sampling also extend to images with varying size and spectrum distribution, affording improved reconstruction

Image Reconstruction usingCompressive Sensing Techniques – A Survey

Compressed Sensing (CS) are utilized for reconstructing the images with minimum quantity of instances at a lower rate. Nowadays, compressive sensing is essentially in the regions of computer visualization problems, signal processing, applied Maths, and optical engineering. In biomedical imaging, image reconstruction using CS is becoming very familiar because it is very effective for improving the image quality. The basic principle behind CS is that sparse signals that contains few number of non-zero coefficients that can be effectively exploited and stored in the basis which can be used to recover any lost signal. Once we construct a signal using CS, reconstruction is also needed. In the existing literature, the various types of reconstruction algorithms are used in the region of CS. This paper gives a review of CS techniques used by researchers for image reconstruction and also a detailed discussion about the applications of CS technique in image processing. The performance of varied forms of CS for image reconstruction is compared based on the final image quality with quantifying parameters such as high Signal-to-Noise Ratio (SNR). Introduction The normal approach of reconstructing signals from processed data observes the Shannon sampling theorem, which notifies that the sampling rate should be the highest frequency twice (i.e., fs ≥ 2fm). Many numbers of samples are required for this approach. Similarly, the basic theorem of linear algebra suggests that the number of measurements of a discrete finite-dimensional signal should be at least as large as its dimension in order to ensure reconstruction. In other words, the above two usual theorems are directly proportional to the number of instances accurately i.e., more sample means more accurate results. But nowadays the invention of a beautiful technique named as CS yields a new approach to reconstruct signals using a minimum number of instances at lower rate. CS also helps to solve the image processing and computer visualization problems.

High Resolution Image Reconstruction Using Fast Compressed Sensing Based on Iterations

WSEAS Transactions on Signal Processing archive, 2018

As a powerful high resolution image modeling technique, compressive sensing (CS) has been successfully applied in digital image processing and various image applications. This paper proposes a new method of efficient image reconstruction based on the Modified Frame Reconstruction Iterative Thresholding Algorithm (MFR ITA) developed under the compressed sensing (CS) domain by using total variation algorithm. The new framework is consisted of three phases. Firstly, the input images are processed by the multilook processing with their sparse coefficients using the Discrete Wavelet Transform (DWT) method. Secondly, the measurements are obtained from sparse coefficient by using the proposed fusion method to achieve the balance resolution of the pixels. Finally, the fast CS method based on the MFR ITA is proposed to reconstruct the high resolution image. In addition, the proposed method achieves good PNSR and SSIM values, and has shown faster convergence rate when performed the MFR ITA un...

Compressive Sensing based Image Compression and Recovery

Compressive sensing is a new paradigm in image acquisition and compression. The CS theory promises recovery of images even if the sampling rate is far below the nyquist rate. This enables better acquisition and easy compression of images, which is more advantageous when the resources at the sender side are scarce. This paper shows the CS based compression and two recovery two methods i.e., l1 optimization and TSW CS recovery. Experimental results show that CS provides better compression, and TSWCS provides better recovery with less relative error recovery than l1 optimization. It is also observed that use of increased measurements leads to reduced error.

A Novel Image Compressive Sensing Method Based on Complex Measurements

2011 International Conference on Digital Image Computing: Techniques and Applications, 2011

Compressive sensing (CS) has emerged as an efficient signal compression and recovery technique, that exploits the sparsity of a signal in a transform domain to perform sampling and stable recovery. The existing image compression methods have complex coding techniques involved and are also vulnerable to errors. In this paper, we propose a novel image compression and recovery scheme based on compressive sensing principles. This is an alternative paradigm to conventional image coding and is robust in nature. To obtain a sparse representation of the input, discrete wavelet transform is used and random complex Hadamard transform is used for obtaining CS measurements. At the decoder, sparse reconstruction is carried out using compressive sampling matching pursuit (CoSaMP) algorithm. We show that, the proposed CS method for image sampling and reconstruction is efficient in terms of complexity, quality and is comparable with some of the existing CS techniques. We also demonstrate that our method uses considerably less number of random measurements.

Fast Compressed Sensing Based High Resolution Image Reconstruction

High Resolution Image Reconstruction with Compressed Sensing based on Iterations

International Journal of Hybrid Information Technology, 2016

This paper proposes a new method of efficient image reconstruction based on the Modified Frame Reconstruction Iterative Thresholding Algorithm (MFR ITA) developed under the compressed sensing (CS) domain by using total variation algorithm. The new framework is consisted of three phases. Firstly, the input images are processed by the multilook processing with their sparse coefficients using the Discrete Wavelet Transform (DWT) method. Secondly, the measurements are obtained from sparse coefficient by using the proposed fusion method to achieve the balance resolution of the pixels. Finally, the fast CS method based on the MFR ITA is proposed to reconstruct the high resolution image. The proposed method achieved superior results on real images, and demonstrate qualitative improvements in terms of PSNR and SSIM values. Furthermore, achieved good reconstruction SNR in the presence of noise.

A Compressed Sensing Approach to Image Reconstruction

IJSRD, 2013

compressed sensing is a new technique that discards the Shannon Nyquist theorem for reconstructing a signal. It uses very few random measurements that were needed traditionally to recover any signal or image. The need of this technique comes from the fact that most of the information is provided by few of the signal coefficients, then why do we have to acquire all the data if it is thrown away without being used. A number of review articles and research papers have been published in this area. But with the increasing interest of practitioners in this emerging field it is mandatory to take a fresh look at this method and its implementations. The main aim of this paper is to review the compressive sensing theory and its applications.

Measurement-Adaptive Sparse Image Sampling and Recovery (original) (raw)

Related papers