Maximum a posteriori video super-resolution with a new multichannel image prior

Faculty of Physics, Department of Electronics, Computers, Telecommunications and Control, National and Kapodistrian University of Athens, Panepistimiopolis, Zografos, 15784 Athens, Greece.
IEEE Transactions on Image Processing (Impact Factor: 3.63). 06/2010; 19(6):1451-64. DOI: 10.1109/TIP.2010.2042115
Source: PubMed


Super-resolution (SR) is the term used to define the process of estimating a high-resolution (HR) image or a set of HR images from a set of low-resolution (LR) observations. In this paper we propose a class of SR algorithms based on the maximum a posteriori (MAP) framework. These algorithms utilize a new multichannel image prior model, along with the state-of-the-art single channel image prior and observation models. A hierarchical (two-level) Gaussian nonstationary version of the multichannel prior is also defined and utilized within the same framework. Numerical experiments comparing the proposed algorithms among themselves and with other algorithms in the literature, demonstrate the advantages of the adopted multichannel approach.

Download full-text


Available from: Stefanos P. Belekos,
92 Reads
  • Source
    • "Protter et al. had come up with a simple super resolution technique by applying denoising method called Nonlocal-Means algorithm (NLM) [11]. Other approaches to the SR problems include Maximum a posteriori (MAP) [12], iterative back-projection [13], POCS [2], [14], [15] and wavelet-based algorithm [16]. Meanwhile, learning-based SR algorithms work by extracting redundant high-frequency image information from training samples which contained known HR components. "
    [Show abstract] [Hide abstract]
    ABSTRACT: Generating high-resolution images or videos have become an essential need for digital image processing and analysis especially in the forensic field. Compressed and at low resolution video frames of common security surveillance videos are found to be very low in clarity and degraded with many noises, distortions, blurs, bad illumination and video compression artifact. This could interfere during image interpretation and analysis process. Using super resolution methods, high resolution image is obtained from a set of low resolution images, after it had undergone two main processes; image registration process based on Keren algorithm and image reconstruction process based on Projection onto Convex Set (POCS) on frequency domain. The validation process of output is done by calculating the Mean Square Error (MSE) and Peak Signal to Noise Ratio (PSNR) value to show the comparison of image quality. The proposed combination methods were evaluated against other typical resampling methods and sparse representation method. The experimental results had shown that the sparse representation method has the highest PSNR mean value but our proposed combinatorial method is comparable to it as the difference of the mean is too small.
  • Source
    • "Iteration back projection (IBP) [3], [4] algorithms estimate the HR image by iteratively back projecting the error between simulated LR images and the observed ones. Maximum a posteriori (MAP) [5]–[7] approaches adopt the prior probability of target HR images to stabilize the solution space under a Bayesian framework. Projection on convex sets (POCS) [8], [9] tends to consider the solution as an element on a convex set defined by the input LR images. "
    [Show abstract] [Hide abstract]
    ABSTRACT: Super-resolution technology provides an effective way to increase image resolution by incorporating additional information from successive input images or training samples. Various super-resolution algorithms have been proposed based on different assumptions, and their relative performances can differ in regions of different characteristics within a single image. Based on this observation, an adaptive algorithm is proposed in this paper to integrate a higher level image classification task and a lower level super-resolution process, in which we incorporate reconstruction-based super-resolution algorithms, single-image enhancement, and image/video classification into a single comprehensive framework. The target high-resolution image plane is divided into adaptive-sized blocks, and different suitable super-resolution algorithms are automatically selected for the blocks. Then, a deblocking process is applied to reduce block edge artifacts. A new benchmark is also utilized to measure the performance of super-resolution algorithms. Experimental results with real-life videos indicate encouraging improvements with our method.
    IEEE Transactions on Image Processing 04/2012; 21(3-21):1031 - 1045. DOI:10.1109/TIP.2011.2166971 · 3.63 Impact Factor
  • Source
    • "For example, it is difficult to incorporate the prior information about HR images using frequency domain methods. For this reason, many spatial domain reconstruction methods have been developed in recent decades, including the nonuniform interpolation approach [9], [10], iterative back-projection approach [11], [12], projection onto convex sets approach [13], [14], maximum likelihood approach [15], maximum a posteriori (MAP) approach [16], [17], joint MAP approach [18]–[21], and the hybrid approach [22]. Recently, some SR algorithms without Fig. 1. "
    [Show abstract] [Hide abstract]
    ABSTRACT: Total variation (TV) has been used as a popular and effective image prior model in regularization-based image processing fields, such as denoising, deblurring, super-resolution (SR), and others, because of its ability to preserve edges. However, as the TV model favors a piecewise constant solution, the processing results in the flat regions of the image being poor, and it cannot automatically balance the processing strength between different spatial property regions in the image. In this paper, we propose a spatially weighted TV image SR algorithm, in which the spatial information distributed in different image regions is added to constrain the SR process. A newly proposed and effective spatial information indicator called difference curvature is used to identify the spatial property of each pixel, and a weighted parameter determined by the difference curvature information is added to constrain the regularization strength of the TV regularization at each pixel. Meanwhile, a majorization–minimization algorithm is used to optimize the proposed spatially weighted TV SR model. Finally, a significant amount of simulated and real data experimental results show that the proposed spatially weighted TV SR algorithm not only efficiently reduces the “artifacts” produced with a TV model in fat regions of the image, but also preserves the edge information, and the reconstruction results are less sensitive to the regularization parameters than the TV model, because of the consideration of the spatial information constraint.
    IEEE Transactions on Circuits and Systems for Video Technology 03/2012; 22(3):379-392. DOI:10.1109/TCSVT.2011.2163447 · 2.62 Impact Factor
Show more