A fast scheme for image size change in the compressed domain

Abstract

Given a video frame in terms of its 8 8 block-DCT coefficients, we wish to obtain a downsized or upsized version of this frame also in terms of 8 8 block-DCT coefficients. The DCT being a linear unitary transform is distributive over matrix multiplication. This fact has been used for downsampling video frames in the DCT domain. However, this involves matrix multiplication with the DCT of the downsampling matrix. This multiplication can be costly enough to trade off any gains obtained by operating directly in the compressed domain. We propose an algorithm for downsampling and also upsampling in the compressed domain which is computationally much faster, produces visually sharper images, and gives significant improvements in PSNR (typically 4-dB better compared to bilinear interpolation). Specifically the downsampling method requires 1.25 multiplications and 1.25 additions per pixel of original image compared to 4.00 multiplications and 4.75 additions required by the method of Chang et al. Moreover, the downsampling and upsampling schemes combined together preserve all the low-frequency DCT coefficients of the original image. This implies tremendous savings for coding the difference between the original frame (unsampled image) and its prediction (the upsampled image). This is desirable for many applications based on scalable encoding of video. The method presented can also be used with transforms other than DCT, such as Hadamard or Fourier.

A DCT-Domain Video Transcoder for Spatial Resolution Downconversion

“…This approach was shown to achieve significant performance improvement over the filtering schemes [14]. [8] interpreted the DCT decimation as basis vectors resampling, and presented a compressed-domain approach for the DCT decimation as described below. …”

“…It is, however, very computationally intensive for real-time applications, even though the motion-vectors and coding-modes of the incoming bit-stream can be reused for fast processing. [2][3][4][5][6][7][8][9][10][11][14][15]. In [10], a simplified pixel-domain transcoder (SPDT) was proposed to reduce the computational complexity of the cascade transcoder by reusing motion vectors and merging the decoding and encoding process and eliminating the IDCT and MC (Motion Compensation) operations.…”

“…To achieve the goal of universal multimedia access (UMA), the video contents need to be adapted to various channel conditions and user equipment capabilities. Spatial resolution reduction [5][6][7][8][9] is one of the key issues for providing UMA in many networked multimedia applications. In realizing transcoders, the computational complexity and picture quality are usually the two most important concerns and need to be traded off to meet various requirements in practical applications.…”

Digital Video Transcoding

“…However, compared to SDDT, its flexibility is achieved using additional DCT-MC and frame memory, which results in a significantly higher cost in computation and storage. It is thus often adopted for downscaling applications, where the encoder-side DCT-MC and memory will not cost much, since the encoder operates at a reduced resolution [39], [40]. Table 1 compares the computational complexity of five different MPEG-2 transcoders: CPDT, SDDT, CDDT, CPDT using full-scale full-search motion reestimation (DEC-ENC1), and CPDT using three-step fast search motion reestimation (DEC-ENC2).…”

Quality-aware predictor-based adaptation of still images for the multimedia messaging service

“…Solutions have been proposed to speed up transcoding under constraints such as maximum file size and maximum resolution. Some solutions address the problem of estimating the transcoded file size of a single JPEG image, but these are still computationally intensive (in addition to necessitate extensive modifications to existing JPEG software libraries) or overly rigid, considering, for example only scaling by a power of two because they can be performed efficiently in the transform (DCT) domain [15,24,38]. Other solutions for image transcoding have been proposed in the broader context of Web or low-bandwidth resource access, if some solutions are designed to transcode an image so that it fits the constraints while minimizing transcoding time [16], others use a small number of fixed transcoding profiles setting both compression parameters and maximum image resolution to achieve adaptation [23]; neither quite expressing the problem in terms of explicit maximization of resulting image quality or user experience.…”