중첩웨이브렛 변환과 엔트로피 제한 Lattice 벡터 양자화를 이용한 영상 부호화 = Image Coding using Overlapping Wavelet Transform and Entropy Constrained Lattice Vector Quantization

다국어 초록 (Multilingual Abstract)

This paper presents an image coding technique employing an overlapping wavelet transform(OWT) and an entropy constrained lattice vector quantizer ECLVQ). Wavelet transformed image has properties of multiresolution image coding and in according to human visual system(HVS). In multiresolution image coding, available versions of the original image at different resolution are easily obtained. ECVQ, aiming at the minimizing of the distortion for a fixed entropy of the quantizer output, is well combined with the multiresolution image coding which yields high compression ratio and good image quality. But ECVQ has coding complexity which grows exponentially with the vector dimension and large codebook size. Unlike ECVQ, ECLVQ has lower coding complexity in comparison with ECVQ.
In the proposed technique, the OWT-ed image is formatted into vector in according to the amplitude of energy in the band. Bands of multiresolution decomposed image using a wavelet transform have horizontal, vertical, disgonal components. Then a structured ECVQ, i.e, ECLVQ is employed to encode the formatted vectors. The variable-length code table size, which grows exponentially with the vector dimension and bit-rate are greatly reduced by grouping the similar codes. In a word, vectors which have a same absolute value are grouped in a code vector for reducing codebook size. In addition to Optimal index, scale factor and sign bit are encoded and transmitted by the additional information.
The performance of the proposed technique is evaluated by the computer simulations, and compared with the unstructured ECVQ using wavelet transform and based on the conventional VQ by graph of PSNR vs. bpp. It is found that the OWT-ECLVQ has a few lower performance to the ECVQ using wavelet transform, while the OWT-ECLVQ has an advantage over the ECVQ using wavelet transform in terms of the encoding complexity and the reduction of the blocking effect.

목차 (Table of Contents)

• 목차 = ⅰ
• ABSTRACT = ⅱ
• Ⅰ. 서론 = 1
• Ⅱ. 웨이브렛 변환 = 4
• 2-1. 웨이브렛(Wavelet) 변환 = 4
• 2-2. 웨이브렛 변환과 다해상도 표현 = 6
• 2-3. 다차원 신호로의 확장 = 11
• 2-4. 중복 웨이브렛 변환(OWT) = 12
• Ⅲ. 격자(Lattice) = 18
• 3-1. D_(n), E_(n), Λ_(n) 격자 = 18
• 3-2. 정수 격자(Integer Lattice) = 19
• Ⅳ. 벡터 양자화 = 22
• 4-1. LBG 알고리즘 = 24
• 4-2. LVQ(Lattice Vector Quantization) = 26
• 4-3. ECVQ(Entropy Constrained VQ) = 27
• 4-4. ECLVQ(Entropy Constrained Lattice VQ) = 28
• 4-5. DCT-CVQ(Classified VQ) = 29
• Ⅴ. 구현된 OWT-ECLVQ = 32
• 5-1. OWT를 통한 영상의 다해상도 분할 = 32
• 5-2. 경계정보를 이용한 적응적 벡터 형성 방법 = 34
• 5-3. OWT-ECLVQ = 37
• Ⅵ. 모의실험 및 결과 = 40
• Ⅶ. 결론 = 47
• Ⅷ. 참고문헌 = 48