Justy W. C. Wong

Predictive motion estimation for reduced-resolution video from high-resolution compressed video

To convert a compressed video sequence to a lower-resolution compressed video, one typically needs to decompress the original sequence, down-sample each frame, and recompress it. It involves motion estimation in the reduced sequence which is computational intensive. We propose a novel fast algorithm to predict the motion vector of the reduced video by using the original motion information in the compressed bitstream. We achieve a much higher quality than existing algorithms with low additional complexity.

Data hiding and watermarking in JPEG-compressed domain by DC coefficient modification

JPEG is a common image format in the WWW and can potentially be used to hide data for secure internet communication and watermark for copyright control. In this paper, we propose an algorithm to embed the secret or watermark information. The proposed algorithm, named Watermarking by DC Coefficients Modification (WDCM), assumes that the quality factor used in JPEG compression is known. We observe that it is perceptually undetectable if the DC coefficients in certain texture-rich blocks are modified by a small amount. We thus embed the secret information as a binary bit sequence in the quantized DC coefficients in those texture rich blocks. The watermark embedding process can be applied in compression domain without re-encode the data. The information bits are randomized by some pseudo-random noise (PN) sequences, the keys of which are needed for the decoding of the secret information. By embedding the information in the DC components, the proposed algorithm is robust to common JPEG compression if the quality factor is known.

Hybrid inverse halftoning using adaptive filtering

We propose a novel fast inverse halftoning technique using a combination of spatial varying filtering and spatial invariant filtering. The proposed algorithm is significantly simpler than most existing algorithms. Without explicit edge region classification, the proposed spatial varying filter is shown to be capable of preserving edges effectively.

Fast time scale modification using envelope-matching technique (EM-TSM)

Time scaling of speech and audio signals is one of the key features of the upcoming MPEG4 standard. Synchronized Overlap-and-Add (SOLA) is a time scaling algorithm known to achieve good speech quality. One problem of SOLA is that it requires a large amount of computation in the search of the best matching point between the analysis and synthesis frames. This is especially serious when multiple time scaling factors are to be supported as in the case of MPEG4. In this paper, we propose a technique called envelop-matching to simplify the computation with effectively the same quality. In envelop-matching, zero crossing locations are used as features for the search. We propose a very fast algorithm for the distortion computation.

Image watermarking using spread spectrum technique in log-2-spatio domain

In this paper, we propose to embed the watermark information in the log-2-spatio domain by means of spread spectrum technique. In log-2-spatio domain, the variance of the information is reduced significantly. This improves the efficiency and robustness of spread spectrum technique. Low intensity and mid-band regions are selected to embed the information in order to guarantee an invisible watermark as well as the robustness to JPEG compression. Simulation results show that the embedded information still survives up to the JPEG compression ratio of 14.7.

On improving the intelligibility of synchronized over-lap-and-add (SOLA) at low TSM factor

We propose an algorithm to modify the synchronized overlap-and-add (SOLA) technique. SOLA is a popular technique for time scale modification of speech and audio signal. It changes the time scale of the signal while maintaining the pitch information. However, when the time scale of the signal is compressed more, the synthesized speech of SOLA has high articulations rate that it becomes almost impossible for the human ear to comprehend. The proposed algorithm can enhance the quality if the time scaled signal by using a time varying time scale factor which is locally adapted to the input signal. It is found that the intelligibility of the synthesized signal using the proposed algorithm is considerably better than the one synthesized from simple SOLA.

Modified predictive motion estimation for reduced-resolution video from high-resolution compressed video

To convert a compressed video sequence to a lower-resolution compressed video, one typically needs to decompress the original sequence, down-sample each frame, and recompress it. It involves motion estimation in the reduced sequence which is computational intensive. Predictive motion estimation was proposed to generate the motion vectors of the reduced video by using the original motion information in the compressed bitstream. However, it was described only for the case of 4 to 1 down-conversion. In this paper, we propose a modified predictive motion estimation to work on other down-conversion ratios at lower complexity.

Reducing computational complexity of dynamic time warping-based isolated word recognition with time scale modification

In this paper, we propose an algorithm to reduce the computational complexity of dynamic time warping for isolated speech recognition. Prior to the feature extraction process in speech recognition, we apply time scale compression both on the reference and testing utterances. Experimental results show that the computational complexity can be reduced by up to 75% without affecting the accuracy of recognition. The time scale process can also suppress the effect of noise to a certain degree so that the recognition accuracy can be improved for noisy test utterances. The proposed algorithm can solve the problem of high mismatch of utterance duration between two utterances.

Fast motion estimation for video resolution down-conversion using spatial-variant filter

To convert a compressed video sequence to a lower-resolution compressed video, one typically needs to decompress the original sequence, down-sample each frame, and recompress it. It involves motion estimation in the reduced sequence, which is computational intensive. In this paper, a novel fast motion estimation algorithm is proposed to predict the motion vectors of the reduced-resolution video without performing any search. The motion vectors are predicted by applying spatial-variant filters to the motion vectors of the original compressed high-resolution video. In simulations, our method outperforms other existing fast predictive algorithms which involve no searching.

An adaptive video sub-sampling technique for the conversion between high and low resolution

An adaptive sub-sampling technique is proposed for the conversion of high-resolution video to low-resolution video, which will eventually be converted back to high resolution. This technique can preserve high spatial frequency information by exploiting the temporal correlation of stationary regions without any extra storage or transmission bandwidth. Coding complexity can be reduced by a factor of four if the video is to be eventually displayed in high-resolution, with small degradation in visual quality and peak-signal-to noise ratio. Most of the edge information can be preserved when the low-resolution video is converted back to high-resolution video and significant improvement can be obtained compared with conventional methods.