Fast Image Restoration for Reducing Block Artifacts Based on Adaptive Constrained Optimization
✍ Scribed by Tae Keun Kim; Joon Ki Paik
- Publisher
- Elsevier Science
- Year
- 1998
- Tongue
- English
- Weight
- 426 KB
- Volume
- 9
- Category
- Article
- ISSN
- 1047-3203
No coin nor oath required. For personal study only.
✦ Synopsis
In this paper we propose a fast adaptive image restoration filter using DCT-based edge-classification for reducing block [2].
artifacts in compressed images. In order to efficiently reduce
To reduce block artifact without increasing the bit rate, block artifacts, edge direction of each block is classified by the following approaches have been proposed: (i) lowpass using the DCT coefficients, and an adaptive constrained least filtering on the boundary region between blocks [3]; (ii) squares (CLS) filter along the edge direction is used for filtering iterative image restoration based on the theory of projecthe corresponding block. The proposed research is based on tions onto convex sets (POCS) [2, 4] or regularization [5]; the observation that degradation due to block artifacts is a (iii) prediction of AC coefficients using the mean squared nonlinear and many-to-one mapping operator. Then, we prodifference of slope (MSDS) between the neighboring pixels pose an approximated version of a constrained optimization on their boundaries [6]; and (iv) optimization based on technique as a restoration process for removing the nonlinear boundary orthonormal function method [7, 8]. The lowpass and space-varying degradation operator. For real-time applicafiltering approach has the advantage that it requires neither tion, the proposed restoration filter is implemented in the form of a truncated FIR filter, which is suitable for postprocessing additional information to be transmitted nor significant the images in real-time video systems such as HDTV or video computation overhead. But this approach results in unnecconferencing systems. © 1998 Academic Press essary image blurring, which critically degrades the image quality. The method based on POCS or regularization needs significantly increased amount of processing time
1. Introduction
due to iterative procedure. So it is not suitable for realtime video processing. AC prediction method has the dis-As the demand for video communication has grown, advantage of high computation complexity for optimizamany efficient image compression methods have been detion process, and boundary orthonormal function based veloped and standardized. Especially, high quality image optimization method may result in different type of distorcommunication with low bit-rate is gaining growing intertion at the cost of reducing block artifacts. ests in applications to video conferencing, videophone, in-In order to solve the above mentioned problems, we teractive TV, etc.
propose a fast adaptive image restoration filter to effi-The block discrete cosine transform (BDCT) is the most ciently reduce block artifacts in BDCT-based comwidely used technique for the compression of both still and pressed images. moving images. Several international standards, including
The outline for the remaining part of this paper is as JPEG for still images, H.261 for videophone/teleconferfollows. In Section 2, a new degradation model for block ence, MPEG-1 for moving pictures and associated audio artifacts is described and formulation of the proposed in digital storage media such as video-CD, and MPEG-2 adaptive image restoration based on the degradation for full-digital HDTV, commonly use the BDCT due to model is given in Section 3. The FIR implementation of the its high energy compaction property [1].
proposed algorithm for real-time processing is proposed in The main drawback of the BDCT based compression Section 4. And experimental results and conclusions are techniques is, however, block artifacts which results from respectively given in Sections 5 and 6. the independent processing of the blocks without taking
2. DEGRADATION MODEL FOR BLOCK ARTIFACT
* This research was supported by Korea Institute of Information Tech-In order to simplify derivation of the proposed algonology Assessment as a part of University Basic Research Support project. rithm, we use the following, one-dimensional vector repre-234