Stmicroelectronics S.r.l.

Luca Pezzoni

Danilo Pietro Pau

Emiliano Mario Piccinelli

Daniele Bagni

Fabrizio Simone Rovati

Davide Giovenzana

Daniele Sirtori

A method for coding video data according to the MPEG-2 standard is provided. Non-Intra fields are identified among Intra fields and various coding options exist for the non-Intra fields, including removal of temporal redundancy using a motion estimation algorithm and identification of predictor macroblocks for providing approximation according to a prediction mode selectable among respective predicting modes of the different types of non-Intra fields. The method includes calculating a discrete cosine transform (DCT) for blocks of data of the predictor macroblocks according to a frame mode of decomposition with the blocks being composed of lines of data belonging to even lines semifield and to odd lines semifield, or in a field mode of decomposition with the blocks being composed of lines of data belonging to the same semifield. The data resulting from the DCT carried out by blocks of data to be stored is quantized and coded. The efficiency of the method for coding video data according to the MPEG-2-standard is increased by selecting between field or frame modes and motion compensation or no motion compensation with an algorithm that assesses and compares significant complexity indexes of the macroblocks.

Video pictures compression and coding with decision step for field/frame and motion vectors DCT

A method for compressing encoding data of a sequence of pictures is based on a motion estimation among the successive images to remove the temporal redundancy from the data. The method recognizes a 3:2 pulldown conversion of a series of photograms of a filmed sequence in a sequence of TV frames. The TV frames have a number greater than the number of the photograms by duplicating certain pictures in a certain order. The redundancy due to such picture duplications is eliminated.

Detection of a 3:2 pulldown in a motion estimation phase and optimized video compression encoder

A method for recognizing a progressive or an interlaced content of video pictures during their processing in a coder includes performing a number of operations on at least one of the luminance or chrominance components of the video signal. A macroblock belonging to a frame of a preceding picture is defined, and a first pair of coefficients on the selected luminance or chrominance component of the video signal is calculated. A first counter is incremented at each positive verification when one of the coefficients is greater than the other coefficient by a determined amount. A second counter is incremented at each macroblock being tested. A second pair of coefficients is calculated for each row of each Top semi-frame. A third counter is incremented at each positive verification when one of the coefficients is greater than the other coefficient by a determined amount. A fourth counter is incremented at each row tested. Verification is made as to whether a content of the first counter is greater than a content of the second counter by a determined amount, and whether, at a same time, a content of the third counter is greater than a content of the fourth counter by a determined amount. If so, then the frame composed of the Top and Bottom semi-frames of the current picture is an interlaced frame, and if not, then the frame is a progressive frame.

Method for recognizing a progressive or an interlaced content in a video sequence

Through a comparison of the values of pixels of a first candidate predictor macroblock having pixels in homologous positions to those of a reference macroblock of identical position on the frame to that of the macroblock being estimated on a reference frame of the present sequence of picture frames, a pre-established cost function is evaluated for each comparison. The best predictor is the one producing the minimum value of cost function. The comparison may even include the summing to each predictor candidate of an update vector of smaller dimensions than the macroblocks, chosen among a plurality of pre-established update vectors, for accelerating the convergence process of the comparison.

Spatio-temporal recursive motion estimation with 1/2 macroblock and 1/4 pixel undersampling

The compression and coding of digital data pertaining to video sequences of pictures including motion estimation for removing temporal redundance are provided by recognizing the occurrence of a change of a scene to control the prediction computation of the pictures. This control is provided using a forward motion estimation for pictures preceding the change of scene, and using a backward motion estimation for pictures subsequent the change of scene of a given sequence. A change of a scene is reliably detected by checking two distinct indexes. These indexes are based on an average value of a smoothness index of the preestablished number of last processed pictures. The smoothness index of a motion field of each picture is calculated by analyzing the motion vectors for all the macroblocks of a subdivision of the picture, except for the peripheral macroblocks. Spurious detections in the event of noisy pictures, zooming and other situations that may adversely affect either one of the indexes are prevented.

Growing community of inventors

Cormano, Italy

Luca Pezzoni