Reduced-reference image quality metric based on statistic model in complex wavelet transform domain

Image Reconstruction from Multiscale Singular Points Based on the Dual-Tree Complex Wavelet Transform

“…e inverse Gaussian distribution (IGD) [30] has fewer errors and parameters than the MRD and GMM, so an adaptive IGD model can be applied to fit the probability density function (PDF) of the DT-CWT coefficient magnitude based on the Kullback-Leibler divergence (KLD) [15]. e GMM, MRD, and IGD are expressed, respectively, as…”

“…Selected from the USC-SIPI image database [26], the test images were the grayscale "House" image of size 256 × 256 [as shown in Figure 6a] and the grayscale "Boat," "Elaine," "Barbara" and "Peppers" of size 512 × 512 (as shown in the first row of Figure 7). To measure the quality of the reconstructed images, as well as the PSNR defined in Equation (15), two other image quality assessment criteria were considered, namely.…”

“…In view of the aforementioned drawbacks of DWT in IR from SPs, this paper proposes a model for SP selection in the domain of dual-tree complex wavelet transform (DT-CWT). DT-CWT is widely used in image processing, such as image quality assessment [15], image denoising [16], image watermarking [17], and IR [18]. While retaining multiresolution characteristics and local time-frequency analysis capabilities, DT-CWT also has shift invariance and good directional selectivity [19].…”

Reducing Dimensions of the Feature Vector of an Image Based on Blocking-DCT

“…A summary of these methods and the required equations are listed in Table I, with the number of necessary values should be sent from a sender to the receiver. In following, the rank order correlation between subject qualification and objective measurement, Spearman Rank Order Correlation Coefficient (SROCC) as a suitable performance to compare experiment results of the proposed idea besides the other two methods, RRPSV [7] and FMRP [8] on three famous database LIVE [9], CSIQ [10] and TID2013 [11] would be presented in Table II. Article [7] compared the entropy of the reflective velues of the degraded and the corresponded reference image blocks in size w × w × 3 on the eigenvectors of covariance matrix of all the reference image blocks.…”

“…The eigenvectors matrix in size w × w × 3 should be transmitted besides one entropy value for each block. Article [8] is based on DT-CWT statistics and machine learning for regression. The required calculation time of feature vector for four methods are shown in the last row of Table II according to using a PC with 3.7 GHz Intel Core i7 CPU and 16 GB RAM.…”