ASSTT. PROFESSOR SUBMIITED BY : MEENAKSHI ME-ECE (2nd YEAR) ROLL NO. -13-713 UIET (P.U), CHD.
Imaging is a broad field which covers all aspects of the
analysis, modification, compression, visualization, and generation of images. Image reconstruction can be seen as the solution of a mathematical inverse problem in which the cause is inferred from the effect. As a consequence, measurement and recording techniques designed to produce the images depend deeply on the application considered.
Compressive sensing is a new type of sampling theory,
which predicts that sparse signals and images can be reconstructed easily . CS relies on the empirical observation that many types of signals or image can be well-approximated by a sparse expansion in terms of a suitable basis, that is, by only a small number of non-zero coefficients to be incomplete information. A compression is obtained by simply storing only the largest basic coefficients. When reconstructing the signal the non-stored coefficients are simply set to zero
The problem defined here is image reconstruction using
compressive sensing technique. For that purpose, different techniques and parameter can be implemented and used. The image at the decoder side can be recovered easily with some sparse information , thereby saving the storage memory, fewer measurements and short scan time, reduce data rate and low complexity encoders. Sparsity expresses the idea that the information rate of a continuous time signal may be much smaller than suggested by its bandwidth and CS exploits the fact that many natural signals are sparse or compressible in the sense that they have concise representations when expressed in the proper basis. Incoherence extends the duality between time and frequency and expresses the idea that objects having a sparse representation in one domain must be spread out in the domain in which they are acquired
COMPRESSIVE SENSING
Compressed sensing relies on L1 techniques, which several other
scientific fields have used historically. In statistics, the least squares method was complemented by the norm, which was introduced by Laplace. It was used in matching pursuit in 1993 and basis pursuit in 1998. There were theoretical results describing when these algorithms recovered sparse solutions, but the required type and number of measurements were sub-optimal and subsequently greatly improved by compressed sensing. At first glance, compressed sensing might seem to violate the sampling theorem, because compressed sensing depends on the sparsity of the signal in question and not its highest frequency but this is a misconception .
Sparse
signals with high frequency components
can be highly under-sampled using compressed sensing compared to classical fixed-rate sampling. Compressive Sensing was employed by D.L. Donoho for image reconstruction with different algorithms of recovery for better storage and less CPU recovery time. It can be empolyed with different techniques and their modifications for achieving better PSNR, lesser CPU time w.r.t number of measurements and number of blocks.
Different intraprediction modes or intraprediction
modes can be considered concerning the smoothing and boundary strength and various smoothing filters like butterworth ,adaptive and kalman filters can be used. The effect of noise is not considered with every minimizing technique, which is a crucial obstruction in recovery. While collecting the coefficients (sampling) can be based on other side information too like contrast, number of rows, permutations of the coefficients, strength of pixel can be used. Earlier work has used only variance and energy.
The
constraint of TV minimization can be
wavelet constraint, DCT constraint, radon transform and norms can be considered along with TV. Earlier work has rarely used domains like contourlet, curvelet ,DDWT. For OMP, various side information, variable index set at each iterations, adaptive filtering with the technique, sampling of residual coefficients can be employed.
Implementation of block compressive sensing and applying
transform like wavelet transform (Haar, Shear, Daubechies,), DDWT (Dual tree Discreet Wavelet Transform), radon transform, fourier transform in domains like spatial, contourlet, curvelet etc. Design of Measurement matrix [15].TV minimization either by intraprediction modes , gradient descent method, norm method or with side information and other constraints like DCT, Contourlet [16]. Sampling optimization like adaptive (based on energy, variance , permutations of coefficients, kroneckor product). OMP with side information, applied only on high level coefficients, with generalised form followed by smoothing filters. Analysis of parameters like PSNR, Quantization noise with respect to number of measurements, number of blocks.