Vasant Patil

A Fast Arbitrary Factor Video Resizing Algorithm

We present a new algorithm for resizing video frames in the discrete cosine transform (DCT) space. We demonstrate that a frame resizing operation can be represented as multiplication by fixed matrices and propose a computation scheme which is applicable to any DCT-based compression method. The proposed approach is general enough to accommodate resizing operations with arbitrary factors conforming to the syntax of 16times16 macroblocks. The approach is shown to possess significant computational gain over the faster known state of the art algorithms while achieving similar picture quality

An Arbitrary Frame-Skipping Video Transcoder

In video transcoding, pre-encoded frames may be arbitrarily dropped to freely adjust the video to meet the network and client requirements. Since transcoding is carried out in real-time, incoming motion vectors are reused to reduce the transcoding latency. In this paper, we propose a new motion vector composition scheme for arbitrarily dropping any frame from incoming video bit-stream comprising I, B and P frames. The transcoded bit-stream retains the I-B-P frame structure. Experimental results are presented and compared to show the efficacy of the proposed scheme

An Efficient Motion Vector Composition Scheme for Arbitrary Frame Down-Sampling Video Transcoder

A new motion vector composition scheme, which we call spatial dominant vector selection (SDVS), for arbitrary down-sampling of precoded video using cascaded pixel domain transcoder is presented. The SDVS uses the existing relationship between pixels of a frame and its reference frame(s) via motion vectors and selects a dominant motion vector from a set of candidate vectors to minimize prediction errors. Compared with the existing methods, experimental results show performance gain, in terms of both quality and bit rate, with reduced computation. We also propose a refinement method that adaptively adjusts the search range to further improve the performance at marginal cost

An Effective Motion Re-estimation in Frame-Skipping Video Transcoding

A new motion vector (MV) composition method, which we call temporal bi-directional dominant vector selection (TBDVS), for frame-skipping transcoding of pre-coded video is presented. The TBDVS utilizes a existing relationship, in pre-coded video, between the pixels of a frame and its reference frame(s) via MVs and selects a dominant MV from a set of candidate MVs to minimize prediction errors. It is general enough to handle frame-skipping, including I frame, in an arbitrary manner from a generic I-B-P frame structure video. Experimental results using cascaded pixel-domain transcoder show performance gain over the existing methods in terms of both quality and bit rate. We also propose a refinement method that adaptively adjusts the search range to further improve the performance at marginal cost

A Fast Arbitrary Factor H.264/AVC Video Re-Sizing Algorithm

An algorithm for re-sizing H.264/AVC video in DCT space is presented. We demonstrate that a frame re-sizing operation can be represented as multiplication by fixed matrices and propose an efficient computation scheme. The proposed approach is general enough to accommodate re-sizing operations with arbitrary factors conforming to the syntax of H.264 video. It shows a good PSNR than the spatial domain bilinear and bicubic spline interpolation at much reduced cost.

A DCT domain frame-skipping transcoder

In video transcoding, pre-encoded frames may be dropped from incoming sequence to freely adjust the video to meet network and client requirements. When some of the reference frames are dropped, quantized DCT coefficients of residual signal of non-skipped frames become invalid as they may refer to the reference frame(s) which no longer exists in the transcoded bit-stream. So, it becomes necessary to re-compute the new set of quantized DCT coefficients with respect to past reference frame(s) that acts as reference in the transcoded bit-stream. The pixel-domain approach to this problem involve high computation and quality degradation due to double re-encoding. In this paper, we propose a method to compute the quantized DCT coefficients for non-skipped frames entirely in the DCT-domain. Experimental results are presented and compared.

A Fast Arbitrary Down-Sizing Algorithm for Video Transcoding

We propose a fast algorithm to down-size video frames, by arbitrary factor, directly in DCT domain. The existing methods treat each 8×8 block as a fundamental unit, therefore, involve high cost of reconstructing down-sized frames. We demonstrate that a basic operation in reconstructing 16×16 macroblock, as a whole, can be represented as multiplication by fixed matrices and the computations can be greatly simplified through decomposition of DCT/IDCT matrix operations. Experimental results using cascaded DCT domain transcoder (CDDT) show substantial reduction in cost of reconstructing down-sized frames at quality similar to the existing methods.

DCT domain transcoding of H.264/AVC video into MPEG-2 video

As the number of different video compression standards increase, there is a growing need for conversion between video formats coded in different standards. H.264/AVC is a newly emerging video coding standard which achieves better video quality at reduced bit rate than other standards. The standalone media players that are available in the market do not support H.264 video playback. In this paper, we present novel techniques that can achieve conversion of pre-coded video in H.264/AVC standard to MPEG-2 standard directly in the compressed domain. Experimental results show that the proposed approach can produce transcoded video with quality comparable to the pixel-domain approach at significantly reduced cost.

A Fast Inverse Motion Compensation Algorithm for DCT-Domain Video Transcoder

The existing methods to perform motion compensation (MC) in discrete cosine transform (DCT) treat each 8 times 8 block as a fundamental unit and, therefore, involve the high cost of reconstructing prediction frames, especially when the half-pixel motion vectors (MVs) are involved. The proposed method operates on a block of variable size 16 Ny times 16 Nx, where Nx and Ny are the number of adjoining 16 times 16 macroblocks with common MV, along the horizontal and vertical directions, respectively. We demonstrate that a basic operation in reconstructing an 16 Ny times 16 Nx DCT-MC block, both with integer and half-pixel MV, can be represented as a multiplication by fixed matrices, and the computations can be greatly simplified through decomposition of DCT/ IDCT operations. Experimental results using cascaded DCT domain transcoder show substantial reduction in computations at quality close to pixel domain transcoding.

A generic video transcoder for MPEG streams by arbitrary frame dropping

In video transcoding, the new motion vectors are composed by reusing incoming motion vectors to reduce transcoding latency. Currently available motion vector composition techniques for temporal transcoding work for progressive video comprising I and P-frames. Since I-frames do not carry any motion vector, such techniques do not handle their dropping. In this paper, we propose a motion vector composition scheme for arbitrarily dropping any frame from incoming video bit-stream comprising I, B and P frames. The proposed scheme handles progressive as well as interlaced video. The scheme is shown to work for MPEG videos. To further reduce transcoding latency, we proposed an adaptive motion vector refinement method using a variability measure. The transcoding results are presented and compared with other schemes.