Correcting temporal artifacts in compressive video sampling with motion estimation

Wahidah, Ida; Hendrawan,; Suksmono, Andriyan Bayu; Mengko, Tati L. R.

doi:10.1109/apcc.2013.6766038

Cited by 2 publications

(1 citation statement)

References 7 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The compressive sensing process utilizes a linear measurement scheme with completion using optimization to obtain a subset of the source signal at a rate that is significantly below conventional sampling. The many benefits of CS make it commonly produced in a different form, as including to model traffic on a telecommunications network [6], to compress medical video signal [7], to recover heart sound on telemedicine [8], to compress respiratory audio [9], [10], to compress bio-signal such as EEG or ECG [11], [12], [13], [14], [15] and other applications.…”

Section: Introductionmentioning

confidence: 99%

Lifting Wavelet Transform in Compressive Sensing for MRI Reconstruction

Irawati¹,

Hadiyoso²,

Budiman³

et al. 2020

Journal of Southwest Jiaotong University

View full text Add to dashboard Cite

Compressed sampling in the application of magnetic resonance imaging compression requires high accuracy when reconstructing from a small number of samples. Sparsity in magnetic resonance images is a fundamental requirement in compressed sampling. In this paper, we proposed the lifting wavelet transform sparsity technique by taking wavelet coefficients on the low pass sub-band that contains meaningful information. The application of novel methods useful for compressing data with the highest compression ratio at the sender but still maintaining high accuracy at the receiver. These wavelet coefficient values are arranged to form a sparse vector. We explore the performance of the proposed method by testing at several levels of lifting wavelet transform decomposition, include Levels 2, 3, 4, 5, and 6. The second requirement for compressed sampling is the acquisition technique. The data sampled sparse vectors using a normal distributed random measurement matrix. This matrix is normalized to the average energy of the image pixel block. The last compressed sampling requirement is a reconstruction algorithm. In this study, we analyze three reconstruction algorithms, namely Level 1 magic, iteratively reweighted least squares, and orthogonal matching pursuit, based on structural similarity index measured and peak signal to noise ratio metrics. Experimental results show that magnetic resonance imaging can be reconstructed with higher structural similarity index measured and peak signal to noise ratio using the lifting wavelet transform sparsity technique at a minimum decomposition level of 4. The proposed lifting wavelet transforms and Level 1 magic reconstruction algorithm has the best performance compared to the others at the measurement rate range between 10 to 70. This method also outperforms the techniques in previous studies.

show abstract

Section: Introductionmentioning

confidence: 99%