Baseline JPEG-like DWT CODEC for disparity compensated residual coding of stereo images

Abstract

In this paper we propose a novel stereo image coding technique, which uses an architecture similar to that of a Discrete Cosine Transform (DCT) based baseline JPEG-CODEC [1], but effectively replaces the DCT technology by the more recently popularized Discrete Wavelet Transform (DWT) technology. We show that as a result of this hybrid design, which combines the advantage of two popular technologies, the proposed CODEC has improved rate distortion and subjective image quality performance as compared to DCT based stereo image compression techniques [2,3]. In particular, at very low bit rates (0.15 bpp), we report peak-signal-to-noise-ratio (PSNR) gains of up to 3.66 dB, whereas at higher bit rates we report gains in the order of 1 dB.

A wavelet implementation of the pioneering block-based disparity compensated predictive coding algorithm for stereo image pair compression

“…All test stereo image pairs were of size 512 Â 512. In order to evaluate the performance of the proposed CODEC, we compare its performance against that of the well-known benchmark algorithm of Perkin's (benchmark-1), that uses DCT as the basic compression engine, our previously proposed CODEC of Nayan et al [13] (benchmark-2), which is essentially a waveletblock-based implementation of benchmark-1 and our original DCT-based PBDCPC algorithm of Jiang et al [9,10] (benchmark-3). As parallel axis camera geometry has been used to acquire these images we have limited the PBS along a horizontal direction only.…”

“…Section 3, discuses the DWT-based multiresolution decomposition of a stereo image pair and the formation of wavelet-block-based structure, which forms the foundation for the proposed implementation. Section 4 discusses PBDCPC algorithm's implementation in wavelet domain by first discussing a multi-resolution pioneering block search (MRPBS) procedure and subsequently by re-structuring the resulting predictive error (PE) image and the reference image within a waveletblock structure [13,16] for entropy coding and transmission. In Section 5, we compare the performance of the proposed CODEC with that of traditional DCT-based DCPC schemes [2,15] and a more recently published DWT based DCPC scheme [13] and prove the novel algorithm's improved rate-distortion performance and improved range of performance.…”

“…In parallel to the above-mentioned shift of technology from monocular to binocular vision capability, the popular, base compression technology that was used in almost all image and video coding standards (baseline-JPEG, MPEGs and H26x), discrete cosine transforms [14], are at present being steadily replaced by discrete wavelet transform (DWT)-based technologies that are capable of providing improved rate-distortion performance and additional functionality [1,4,7,[11][12][13][16][17][18]. The culmination of research in this direction resulted in the recent standardization of JPEG-2000 [7] and wavelet domain implementations of MPEG-4 video coding standard.…”

Wavelet Based Stereo Image Pair Coding With Pioneering Block Search

“…At the encoder end, the left image (reference image) and the right image (predicted image) undergo a 3-level dyadic DWT separately and the subband images are subsequently grouped into wavelet blocks such that each block consists of wavelet coefficients from a common spatial location but from different subbands [3][4]. Adapting the technique described in section 2, by replacing each block by a wavelet-block, able the pioneering block search to be performed simultaneously for different subband images.…”

Reversão anaglífica em vídeos estereoscópicos

“…A primeira etapa a ser realizada é a divisão da imagem em macroblocos (cuja dimensão tradicional é 8 x 8 pixels), posteriormente a aplicação de N níveis da DWT é feita sobre os macroblocos. Finalmente os macroblocos resultantes são submetidos uma divisão entre eles e uma constante (THANAPIROM; FERNANDO; EDIRI-SINGHE, 2005) ou por uma matriz com valores positivos (NAYAN; EDIRISINGHE; BEZ, 2002). A abordagem prosposta por Nayan, Edirisinghe e Bez ( 2002), considera a aplicação de três níveis fixos da DWT para cada macrobloco, enquanto a abordagem de Thanapirom, Fernando e Edirisinghe ( 2005) utiliza aplicação de mais de três níveis da DWT para geração de árvores de wavelets, acarretando um aumento no tempo de execução do método.…”