This study intends to establish a sound testing and evaluation methodology based upon the human visual characteristics
for appreciating the image restoration accuracy; in addition to comparing the subjective results with predictions by some
objective evaluation methods. In total, six different super resolution (SR) algorithms - such as iterative back-projection
(IBP), robust SR, maximum a posteriori (MAP), projections onto convex sets (POCS), a non-uniform interpolation, and
frequency domain approach - were selected. The performance comparison between the SR algorithms in terms of their
restoration accuracy was carried out through both subjectively and objectively. The former methodology relies upon the
paired comparison method that involves the simultaneous scaling of two stimuli with respect to image restoration
accuracy. For the latter, both conventional image quality metrics and color difference methods are implemented.
Consequently, POCS and a non-uniform interpolation outperformed the others for an ideal situation, while restoration
based methods appear more accurate to the HR image in a real world case where any prior information about the blur
kernel is remained unknown. However, the noise-added-image could not be restored successfully by any of those
methods. The latest International Commission on Illumination (CIE) standard color difference equation CIEDE2000 was
found to predict the subjective results accurately and outperformed conventional methods for evaluating the restoration
accuracy of those SR algorithms.
We propose a frame-rate conversion algorithm using hybrid-search-based motion estimation (ME) and adaptive motion-compensated interpolation (MCI). The ME method uses three search strategies: recursive search, three-step search with predictions, and single predicted search. One of them, which is best suited for the predicted motion type, is adaptively performed on a block basis. This adaptation process improves the accuracy of the estimated motion vectors without increasing the computational load. With the estimated motion vectors, the proposed MCI method reconstructs high-quality frames, without producing block artifacts, by considering multiple motion trajectories. The method utilizes pixel smoothness constraints besides motion-vector reliability when creating and combining the multiple motion-compensated results to remove block artifacts in regions with unreliable motion vectors. Experimental results show that the proposed ME method produces reliable motion vectors that are closer to true motions. Also, the proposed MCI method achieves better image quality than existing algorithms.
We propose a motion-compensation-based deinterlacing
algorithm using global and representative local motion estimation.
The proposed algorithm first divides an entire image into five regions
of interest (ROIs) according to the temporally predicted motion type
(i.e., global or local) and the spatial position. One of them is for
global motion estimation and the others are for local motion estimation.
Then, dominant motions of respective ROIs are found by adaptive
projection approach. The adaptive projection method not only
estimates dominant local motions with low computational cost, but
also ensures consistent global motion estimation. Using the estimated
motion vectors, adaptive two-field bidirectional motion compensation
is performed. The arbitration rules, measuring the reliability
of motion compensation accurately, produce high-quality
deinterlaced frames by effectively combining the results of motion
compensation and the stable intrafield deinterlacing. Experimental
results show that the proposed deinterlacing algorithm provides better
image quality than the existing algorithms in both subjective and
In color television broadcasting standards, such as National Television System Committee (NTSC) and phase alteration line (PAL), the bandwidth of the chrominance signals are even narrower than those of the luminance signals. Also in digital video standards, the chrominance signals are usually low-pass filtered and subsampled to reduce the amount of data. Because of these reasons, the chrominance signals have poor transition characteristics and the slow transition causes blurred color edges. A color transient improvement algorithm is proposed by exploiting the high-frequency information of the luminance signal. The high-frequency component extracted from the luminance signal is modified by adaptive gains and added to the low-resolution chrominance signals in the proposed algorithm. The gain is estimated to minimize the l2 norm of the error between the original and the estimated pixel values in a local window. The proposed algorithm naturally improves the transient of the chrominance signal as much as that of the luminance signal without overshoots and undershoots. The experimental results show that the proposed method produces steep and natural color edge transition and reconstructs narrow line edges that are not restored by the conventional algorithms.
In broadcast system, the image information is transmitted in the form of luminance and color difference signals. The color difference signals usually undergo blurs by the several reasons and result in smooth transition. It is important for the CTI algorithm not to produce color mismatch in the smooth transition as well as to make the transition sharp. In this paper, the new CTI algorithm which only needs to determine the transition range is proposed. Since the corrected signal does not rely on the high-frequency values, it does not reveal over- and undershoot near edges. To prevent the color mismatch, transition range is found on only one color difference channel. Experimental results show that our algorithm corrects blurred color edges well and is robust to the input images.
Recently, wavelet analysis has been used to denoise a digital image corrupted by noise in the acquisition step. Because of the multiscale decompositions (multiresolution) and translation parameter (locality in space), in addition to the scale parameter of the wavelet there are two types of correlations that can be used to detect edges. Many previous works have exploited the intrascale or interscale dependences alone to denoise the image. In this paper, a denoising method that combines the intrascale and interscale correlations simultaneously is proposed. By manipulation of the wavelet coefficients in successive bands, this noise model is investigated exhaustively and estimated as the well-known Gaussian distribution. With the investigated noise distribution, a new denoising method is proposed. Experimental results show the superiority of the proposed method to the interscale or intrascale method according to objective and subjective criteria.