Deinterlacing algorithm using edge direction from analysis of the DCT coefficient distribution

Sang Jun Park, Gwanggil Jeon, Jechang Jeong

Research output: Contribution to journalArticle

7 Citations (Scopus)

Abstract

We propose a single-field deinterlacing algorithm with compressed domain edge direction information (CD-EDI). This algorithm was verified using the DCT coefficient distributions and pixel domain edge direction information (PD-EDI) in conjunction with an analysis of the neighboring pixels. The CD-EDI, which was calculated in the DCT domain, was first addressed, so that we could categorize the edge direction. We studied the distribution of DCT coefficients in a DCT-encoded block by taking into account the orientations of the four edge directions: 0 (horizontal), π/4 (increase diagonal), π/2 (vertical), and 3π/4 (decrease diagonal). In addition, we studied the PD-EDI, which we acquired by identifying the small pixel variations at the six edge directions. On the basis of an edge-based line average algorithm, a PDEDI was established within the operation window in order to lower the frequency of false judgments about the edge direction where a deinterlacing would be made. Finally, we propose an edge direction confidence (EDC) conception for each missing pixel, which is capable of revealing the accuracy of the edge detection. Since fuzzy sets can be represented by a membership function (MF), we used Gaussian MF to find EDC. The proposed algorithm has a simple EDC-evaluating structure of low complexity, which therefore makes it easy for implementation in hardware. In the extensive simulations conducted for different test sequences, the proposed algorithm outperformed all of the other state-of-the-art deinterlacing methods.

Original languageEnglish
Pages (from-to)1674-1684
Number of pages11
JournalIEEE Transactions on Consumer Electronics
Volume55
Issue number3
DOIs
StatePublished - 2009 Oct 29

Fingerprint

Pixels
Membership functions
Edge detection
Fuzzy sets
Hardware

Keywords

  • Compressed domain video analysis
  • DCT
  • Deinterlacing
  • Edge direction confidence

Cite this

@article{b22d9024997a4a18a09ba2c2ffc8b2a7,
title = "Deinterlacing algorithm using edge direction from analysis of the DCT coefficient distribution",
abstract = "We propose a single-field deinterlacing algorithm with compressed domain edge direction information (CD-EDI). This algorithm was verified using the DCT coefficient distributions and pixel domain edge direction information (PD-EDI) in conjunction with an analysis of the neighboring pixels. The CD-EDI, which was calculated in the DCT domain, was first addressed, so that we could categorize the edge direction. We studied the distribution of DCT coefficients in a DCT-encoded block by taking into account the orientations of the four edge directions: 0 (horizontal), π/4 (increase diagonal), π/2 (vertical), and 3π/4 (decrease diagonal). In addition, we studied the PD-EDI, which we acquired by identifying the small pixel variations at the six edge directions. On the basis of an edge-based line average algorithm, a PDEDI was established within the operation window in order to lower the frequency of false judgments about the edge direction where a deinterlacing would be made. Finally, we propose an edge direction confidence (EDC) conception for each missing pixel, which is capable of revealing the accuracy of the edge detection. Since fuzzy sets can be represented by a membership function (MF), we used Gaussian MF to find EDC. The proposed algorithm has a simple EDC-evaluating structure of low complexity, which therefore makes it easy for implementation in hardware. In the extensive simulations conducted for different test sequences, the proposed algorithm outperformed all of the other state-of-the-art deinterlacing methods.",
keywords = "Compressed domain video analysis, DCT, Deinterlacing, Edge direction confidence",
author = "Park, {Sang Jun} and Gwanggil Jeon and Jechang Jeong",
year = "2009",
month = "10",
day = "29",
doi = "10.1109/TCE.2009.5278042",
language = "English",
volume = "55",
pages = "1674--1684",
journal = "IEEE Transactions on Consumer Electronics",
issn = "0098-3063",
number = "3",

}

Deinterlacing algorithm using edge direction from analysis of the DCT coefficient distribution. / Park, Sang Jun; Jeon, Gwanggil; Jeong, Jechang.

In: IEEE Transactions on Consumer Electronics, Vol. 55, No. 3, 29.10.2009, p. 1674-1684.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Deinterlacing algorithm using edge direction from analysis of the DCT coefficient distribution

AU - Park, Sang Jun

AU - Jeon, Gwanggil

AU - Jeong, Jechang

PY - 2009/10/29

Y1 - 2009/10/29

N2 - We propose a single-field deinterlacing algorithm with compressed domain edge direction information (CD-EDI). This algorithm was verified using the DCT coefficient distributions and pixel domain edge direction information (PD-EDI) in conjunction with an analysis of the neighboring pixels. The CD-EDI, which was calculated in the DCT domain, was first addressed, so that we could categorize the edge direction. We studied the distribution of DCT coefficients in a DCT-encoded block by taking into account the orientations of the four edge directions: 0 (horizontal), π/4 (increase diagonal), π/2 (vertical), and 3π/4 (decrease diagonal). In addition, we studied the PD-EDI, which we acquired by identifying the small pixel variations at the six edge directions. On the basis of an edge-based line average algorithm, a PDEDI was established within the operation window in order to lower the frequency of false judgments about the edge direction where a deinterlacing would be made. Finally, we propose an edge direction confidence (EDC) conception for each missing pixel, which is capable of revealing the accuracy of the edge detection. Since fuzzy sets can be represented by a membership function (MF), we used Gaussian MF to find EDC. The proposed algorithm has a simple EDC-evaluating structure of low complexity, which therefore makes it easy for implementation in hardware. In the extensive simulations conducted for different test sequences, the proposed algorithm outperformed all of the other state-of-the-art deinterlacing methods.

AB - We propose a single-field deinterlacing algorithm with compressed domain edge direction information (CD-EDI). This algorithm was verified using the DCT coefficient distributions and pixel domain edge direction information (PD-EDI) in conjunction with an analysis of the neighboring pixels. The CD-EDI, which was calculated in the DCT domain, was first addressed, so that we could categorize the edge direction. We studied the distribution of DCT coefficients in a DCT-encoded block by taking into account the orientations of the four edge directions: 0 (horizontal), π/4 (increase diagonal), π/2 (vertical), and 3π/4 (decrease diagonal). In addition, we studied the PD-EDI, which we acquired by identifying the small pixel variations at the six edge directions. On the basis of an edge-based line average algorithm, a PDEDI was established within the operation window in order to lower the frequency of false judgments about the edge direction where a deinterlacing would be made. Finally, we propose an edge direction confidence (EDC) conception for each missing pixel, which is capable of revealing the accuracy of the edge detection. Since fuzzy sets can be represented by a membership function (MF), we used Gaussian MF to find EDC. The proposed algorithm has a simple EDC-evaluating structure of low complexity, which therefore makes it easy for implementation in hardware. In the extensive simulations conducted for different test sequences, the proposed algorithm outperformed all of the other state-of-the-art deinterlacing methods.

KW - Compressed domain video analysis

KW - DCT

KW - Deinterlacing

KW - Edge direction confidence

UR - http://www.scopus.com/inward/record.url?scp=70350259883&partnerID=8YFLogxK

U2 - 10.1109/TCE.2009.5278042

DO - 10.1109/TCE.2009.5278042

M3 - Article

AN - SCOPUS:70350259883

VL - 55

SP - 1674

EP - 1684

JO - IEEE Transactions on Consumer Electronics

JF - IEEE Transactions on Consumer Electronics

SN - 0098-3063

IS - 3

ER -