Memory-Efficient Discrete Cosine Transform Domain Weight Modulation Transformer for Arbitrary-Scale Super-Resolution
3
Citation
32
Reference
10
Related Paper
Citation Trend
Abstract:
Recently, several arbitrary-scale models have been proposed for single-image super-resolution. Furthermore, the importance of arbitrary-scale single image super-resolution is emphasized for applications such as satellite image processing, HR display, and video-based surveillance. However, the baseline integer-scale model must be retrained to fit the existing network, and the learning speed is slow. This paper proposes a network to solve these problems, processing super-resolution by restoring the high-frequency information lost in the remaining arbitrary-scale while maintaining the baseline integer scale. The proposed network extends an integer-scaled image to an arbitrary-scale target in the discrete cosine transform spectral domain. We also modulate the high-frequency restoration weights of the depthwise multi-head attention to use memory efficiently. Finally, we demonstrate the performance through experiments with existing state-of-the-art models and their flexibility through integration with existing integer-scale models in terms of peak signal-to-noise ratio (PSNR) and similarity index measure (SSIM) scores. This means that the proposed network restores high-resolution (HR) images appropriately by improving the image sharpness of low-resolution (LR) images.Discrete Cosine Transform (DCT) which has a major role in image and video compression has also a major role in power consumption. Approximate Computing let us trade precision to save power in error resilient applications such as multimedia. Therefore, DCT is a potential candidate for approximation. In this paper, we propose a method for evolutionary design of DCT architecture exploiting the inherent behavior of DCT. Unlike the prior works on DCT approximation, which concentrated mostly on optimizing, replacing, or removing less effective building blocks of DCT, in our proposed method we use the evolutionary method to find new structures for DCT. According to the results, the evolution methods lead to architectures with less area and acceptable accuracy.
Trellis quantization
Cite
Citations (2)
This paper presents a method to accelerate software video encoders by reducing the number of operations for discrete cosine transform (DCT) and quantization. We present a model of DCT coefficients based on the information of motion compensated difference blocks. The relationship between the quantization level and DCT coefficients is also studied. Based on the model, we can adaptively make the decision of DCT calculations. A fast algorithm is also proposed for approximating the 4/spl times/4 low-frequency coefficients. The results show that significant computation reductions can be achieved with negligible peak signal-to-noise ratio (PSNR) degradations.
Trellis quantization
Peak signal-to-noise ratio
Discrete sine transform
Cite
Citations (12)
The performance of version I of the discrete cosine transform(DCT-I) is compared to version II of the discrete cosine transform (DCT-II) on various criteria. The results show that for a Markovian signal with correlation coefficient less than 0.8, the DCT-I performs as well as the DCT-II.
Lapped transform
Discrete sine transform
Cite
Citations (12)
The discrete cosine transform (DCT) is widely applied in various fields, including image data compression, because it operates like the Karhunen-Loève transform for stationary random data. This paper presents a recursive algorithm for DCT with a structure that allows the generation of the next higher order DCT from two identical lower order DCT's. As a result, the method for implementing this recursive DCT requires fewer multipliers and adders than other DCT algorithms.
Discrete sine transform
Lapped transform
Trellis quantization
Cite
Citations (414)
Numerous fast algorithms for the discrete cosine transform (DCT) have been proposed in image and video processing literature. Until recently, it has been difficult to compare different DCT algorithms and select one which is best suited for implementation under a given set of design goals and constraints. In this paper, we propose an approach for design space exploration at the algorithm and behavioral levels using high level synthesis tools. In particular, we study and compare the following nine DCT algorithms: Lee's, Wang's, DIT, DFT, QR, Givens, Arai, MCM, and direct algorithm. The main conclusion of this study is that the best choice among fast DCT algorithms depends on a particular set of design goals and constraints. Another important conclusion is that for almost all sets of implementation goals and constraints more than an order of magnitude improvement can be achieved using algorithm and behavioral design space exploration.
Design space exploration
Algorithm design
Cite
Citations (5)
DCT(Discrete Cosine Transform) 계수 제거 기법은 MC(Motion Compensated)-DCT 기반의 MPEG 비디오에서의 효율적인 율적응 트랜스코딩 기법이다. 그러나, 이들 기법에서는 DCT 계수 제거로 인한 왜곡이 전파되게 되고 종종 심각한 화질 열화를 유발하게 된다. 본 논문에서는 왜곡 전파 특성에 대한 두 가지의 통계적 성질을 제시하고 수식적으로 분석한다. 즉, 현재 프레임의 DCT 계수 제거 왜곡과 이전 프레임에서 전파되어 오는 왜곡간에 상관성이 없음을 보이고 각 프레임의 DCT 제거로 발생되는 전파 왜곡의 누적과 현 프레임의 DCT 계수 제거 왜곡의 합으로 전체 왜곡을 근사할 수 있음을 보인다. 시뮬레이션을 통하여, 본 논문에서 수식적으로 제시한 통계적 특성이 실제 비디오 시퀀스에서 유효함을 실험적으로 증명한다. Discrete cosine transform (DCT) coefficient dropping is well recognized as an efficient rate adaptation transcoding in motion-compensated (MC)-DCT based MPEG-compressed videos. However, in this scheme, the errors incurred by the DCT coefficient-dropping are propagated and often result in significant visual quality degradation. This paper presents two propositions describing well the statistical properties of propagated errors. That is, we propose that the DCT error of the current frame is not correlated to the propagated errors of the previous frames. We also propose that the overall distortions in a given frame can be approximated as the sum of the DCT error of the current frame and the propagated errors from the previous frames. Then, it is shown that several computer simulations with different video sequences verify the effectiveness of the proposed statistical analyses.
Trellis quantization
Transcoding
Cite
Citations (0)
In this paper, an algorithm for image coding based on a sparse 3-dimensional Discrete Cosine Transform (3D DCT) is studied. This is essentially a method for achieving a sufficiently sparse representation using 3D DCT. The simulation results obtained by the algorithm are compared to the 2D DCT. From simulation it is shown that the algorithm, that uses DCT but in 3 dimensions, outperforms the 2D DCT used in JPEG standard and achieves comparable results but has a limitation of line pattern.
JPEG
Lapped transform
Trellis quantization
Discrete sine transform
Cite
Citations (0)
Trellis quantization
Cepstrum
Cite
Citations (1)
Discrete Cosine Transform. Definitions and General Properties. DCT and Its Relations to the Karhunen-Loeve Transform. Fast Algorithms for DCT-II. Two Dimensional DCT Algorithms. Performance of the DCT. Applications of the DCT. Appendices. References. Index.
Discrete sine transform
Lapped transform
Trellis quantization
Discrete Hartley transform
Cite
Citations (1,939)
In this paper, a watermarking method has been proposed based on Discrete Cosine Transform(DCT) which can be used in order to protect copyrighting and to provide right of image ownership. In this method, the original image transferred to DCT domain after dividing into non-overlapped blocks 8×8 and to the same method, watermark image which can be whether a firm mark or any desired image from owner of the art work, after dividing into non-overlapped blocks 4×4, transferred to DCT domain. Watermark image coefficients after one step coding composed with low frequency coefficients of original image and create the final watermark image. On the other hand, the process of reforming watermarked image and extracting the original watermark on the secondary side is extractable by using original image and with reverse mechanism. Experiments show that this method in encountering with a number of routine attacks has a good resistance.
Cite
Citations (0)