Industrial Technology Research Institute

Chia-Wen Lin

Ming-Ting Sun

Jeongnam Youn

Chung-Chieh Kuo

Yu-Chen Lin

Che-Tsung Lin

I-Hsien Lee

Kual-Zheng Lee

Yeh-Kai Chou

Wen-Hao Wang

Hsu-Kai Tsai

Chih-Chung Hsu

Tien-Ying Kuo

Jia-Jyh Luo

Su-Ren Chen

Li-Wei Kang

Yuh-Ruey Lee

Hung-Hseng Hsu

Chih-Sheng Shen

Shih-Ming Hsu

Zhi-Hong Ling

A method and an apparatus for single-image-based rain streak removal are introduced herein. In this method and apparatus, an original image is decomposed into a high frequency part and a low frequency part, and the high frequency image is then decomposed into a rain part and a non-rain part. The non-rain high frequency image and the original low frequency image are used to produce a non-rain image.

Method and apparatus for single-image-based rain streak removal

A method for example-based face hallucination uses manifold learning to project a plurality of training images in a training database and an input low resolution (LR) face image into a same manifold domain, then iteratively refines the reconstruction basis by selecting a training set having k projected training images which best match the parts of the projected LR face image, where k&#x2266;N and N is the number of projected training images. Through the best-match training set, a set of prototype faces are learned, and the set of prototype faces are used as the reconstruction basis to reconstruct a high resolution face image for the input LR face image.

Method and system for example-based face hallucination

A method of selecting a number of candidate prediction modes for a block in a video sequence, the method comprising calculating a cost value of each of prediction modes for each of a predetermined number of blocks, identifying one of the prediction modes having the smallest cost value for the each block, calculating a function value of each of the prediction modes for the each block using a cost function, ranking the prediction modes for the each block by the function value of each of the prediction modes and identifying an ordinal value of the one prediction mode having the smallest cost value, the ordinal value being related to the ordinal number of the one prediction mode after the ranking, calculating a feature value of the each block based on the function value of each of the prediction modes related to the each block, identifying a plurality of sets of blocks, each set of blocks having substantially the same feature value, identifying the number of each set of blocks and calculating a sum of the ordinal value of the one prediction mode for the each block in the each set of blocks, calculating an average value of the sum over the number of the each set of blocks for each set of the plurality sets of blocks, and determining a relation between the average values and the feature values for the predetermined number of blocks.

Methods for selecting a prediction mode

An encoder. A first encoding unit discrete cosine transforms an input frame, quantizes the transformation result, and generates a first frame according to a motion vector. The first encoding unit includes a first feedback unit dequantizing the transformation result, generating a processing signal and a first reconstruction signal according to the dequantization result, and re-quantizing the processing signal to generate a requantization signal. A second encoding unit encodes according to the first reconstruction signal to generate a second frame and an encoding signal. The third encoding unit generates a third frame according to the encoding signal and the re-quantization signal.

Encoders and image encoding methods

An object tracking method may include: receiving frames of data containing image information of an object; performing an object segmentation to obtain an object motion result; and using the object motion result to conduct an object tracking. In particular, the object segmentation may include: extracting motion vectors from the frames of data; estimating a global motion using the motion vectors; and subtracting the global motion from the motion vectors to generate an object motion result.

Object tracking systems and methods utilizing compressed-domain motion-based segmentation

Methods and systems for transcoding a video sequence in a discrete cosine transform (DCT) domain, wherein a transcoder receives a video bit-stream including frames and each of the frames including blocks. The video bit-stream includes an intra-frame and an inter-frame that has been encoded by motion compensation based on the intra-frame or another inter-frame. A DCT-domain motion compensation module in the transcoder re-calculates first DCT coefficients for a target block in the inter-frame. For this re-calculation of the first DCT coefficients, the motion compensation module inputs second DCT coefficients of neighboring blocks in the inter-frame, and calculates partial DCT coefficients, using significant ones of the second DCT coefficients of the neighboring blocks.

Methods and systems for video transcoding in DCT domain with low complexity

An architecture of a fine granularity scalable (FGS) codec has an encoder and a decoder configurable in three prediction modes. The coarse prediction loop in the base layer of the encoder has a switch for selecting either coarse prediction output or fine prediction output in the encoder. The fine prediction loop in the enhancement layer of the encoder also has a switch for selecting either coarse prediction output or fine prediction output. Two-pass encoding is used in the encoder. The first pass extracts coding parameters and classifies macroblocks of a video frame into three groups each being assigned with all-coarse prediction mode, all-fine prediction mode or mix prediction. The second pass uses the assigned modes to encode the macroblocks. A rate adaptation algorithm is provided to truncate the enhancement bit-planes for low bit rate, medium bit rate and high bit rate and allocate bit efficiently for achieving higher video quality.

Method and apparatus for MPEG-4 FGS performance enhancement

A low-complexity spatial downscaling video transcoder and method thereof are disclosed. The transcoder comprises a decoder having a reduced DCT-MC unit, a DCT-domain downscaling unit, and an encoder. The decoder performs the DCT-MC operation at a reduced-resolution for P-/B-frames in an MPEG coded bit-stream. The DCT-domain downscaling unit is used for spatial downscaling in the DCT-domain. After the downscaling and the motion vectors re-sampling, the encoder determines the encoding modes and outputs the encoded bit-stream. Compared with the original CDDT, this invention can achieve significant computation reduction and speeds up the transcoder without any quality degradation.

Low-complexity spatial downscaling video transcoder and method thereof

Methods and systems for providing a video stream from a server to a client over a network include a memory for storing a forward-encoded bit-stream and a reverse-encoded bit-stream for a video data. The forward-encoded bit-stream includes I-frames encoded without inter-frame dependencies and P-frames encoded depending on forward-direction preceding frames, and the reverse-encoded bit-stream includes I-frames and P-frames encoded depending on reverse-direction preceding frames. When the server receives a request with a video cassette recording (VCR) function from the client, the server reads out and transmits frames selectively from among the first, second, third, and fourth frames in accordance with the request. The server can select the closest I-frame to a requested frame in either bit-stream for a fast-mode play or a random-access play, and switch the bit-streams to use subsequent P-frames in a different direction than that of the closest I-frame.

Methods and systems for video streaming with VCR functionality

A multipoint video conferencing system employs a transcoder with a dynamic sub-window skipping technique to enhance the visual quality of the participants of interest. The system firstly identifies the active conferees from the multiple incoming video streams by calculating the temporal and the spatial activities of the conferee sub-windows. The sub-windows of inactive participants are dropped and the saved bits are reallocated to the active sub-windows. Numerous motion vector composition schemes can be used to compose the unavailable motion vectors in the dropped frames due to limited bit-rates or frame-rates of the user clients in video transcoding. The present invention employs a pre-filtered activity-based forward dominant vector selection (PA-FDVS) scheme which provides accurate approximation of motion vectors with low computational cost and memory requirement. Simulation results show the visual quality of the active sub-windows is significantly improved at the cost of degrading the temporal resolution of the inactive sub-windows, which degradation is relatively invisible to human perception.

Video transcoder, video transcoding method, and video communication system and method using video transcoding with dynamic sub-window skipping

A transcoder for transcoding digital video signals includes a decoder and an encoder. In the decoder, an end-of-block (EOB) position of an incoming block received by the decoder is determined and a discrete cosine transform (DCT) block type is determined based on the determined EOB position. A reduced number of DCT coefficients is computed in a subsequent inverse DCT computation based on the DCT block type. In the encoder, if the incoming block is intercoded, no DCT coefficients are computed after the EOB of the incoming blocks is performing a DCT. Further, in the encoder when the incoming block is intercoded, an algorithm is applied to predict which DCT coefficients may become zero after a subsequent quantization operation, and only DCT coefficients that may not become zero are computed in performing the DCT.

Transcoding apparatus and method

A block-based motion-compensated frame interpolation method and apparatus using a block-based video coder operating in low bit rates. Smooth movement of objects between video frames can be obtained without the complexity of pixel-wise interpolation motion estimation that is present in standard motion-compensated frame interpolation (MCI). An additional motion search for interpolating all of the individual pixel trajectories is not required because the interpolation uses block-based motion vector information from a standard codec such as H.26x/MPEG. Video quality is improved by increasing smoothness and the frame rate is increased without a substantial increase in the computational complexity. The proposed block-based MCI method maps from block-wise motion to pixel-wise motion in a motion vector mapping unit. A morphological closure operation and pattern block refinement segmentation of the blocks are provided to close holes in the moving object block and replace the morphologically closed motion block with the most similar pattern selected from a group of 34 patterns. Experimental results show that the visual quality of coded low-bit-rate video can be significantly improved as compared to the frame repetition scheme at the expense of a small increase in the complexity of the decoder.

Fast motion-compensated video frame interpolator

Methods and systems for generating motion vectors for re-encoding video signals are disclosed. The motion vector is determined by the sum of a base motion vector and a delta motion vector. In the case of no frame-skipping, the base motion vector is the incoming motion vector. In the case of frame skipping, the base motion vector is the sum of the motion vectors of the incoming signal since the last re-encoded frame and the current frame. The delta motion vector is optimized by a minimum Sum of the Absolute Difference by searching over a smaller area than if searching for a new motion vector without a delta motion vector. These methods and systems may be used to improve re-encoding digital video signals.

Growing community of inventors

Chiayi, Taiwan

Chia-Wen Lin