Site News  |   Monitor Keywords  |   Monitor Archive  |   Organizer  |   Account Info  |

Moving picture prediction system

Moving picture prediction system description/claims

This application is a Divisional of co-pending application Ser. No. 10/642,508 filed on Aug. 18, 2003 and for which priority is claimed under 35 U.S.C. § 120. The entire contents of the above-identified application is hereby incorporated by reference.

The present invention relates to the prediction of a moving picture implemented, for example, in

a moving picture encoder/decoder used in a portable/stationary video communication device and the like for visual communications in a video telephone system, a video conference system or the like,

a moving picture encoder/decoder used in a picture storage/recording apparatus such as a digital VTR and a video server, and

a moving picture encoding/decoding program implemented in the form of a single software or a firmware as a Digital Signal Processor (DSP).

MPEG-4 (Moving Picture Experts Group Phase-4) Video Encoding/Decoding Verification Model (hereinafter referred to by the initials VM) whose standardization is in progress by ISO/IEC JTC1/SC29/WG11 may be introduced as a conventional type of predictive encoding/decoding in an encoding/decoding system of moving pictures. The VM continues to revise its contents according to the progress being made in standardization of MPEG-4. Here, Version 5.0 of the VM is designated to represent the VM and will be simply referred to as VM hereinafter.

The VM is a system for encoding/decoding each video object as one unit in view of a moving picture sequence being an aggregate of video objects changing their shapes time-/space-wise arbitrarily. FIG. 29 shows a VM video data structure. According to the VM, a time-based moving picture object is called a Video Object (VO), and picture data representing each time instance of the VO, as an encoding unit, is called a Video Object Plane (VOP). If the VO is layered in time/space, a special unit called a Video Object Layer (VOL) is provided between the VO and the VOP for representing a layered VO structure. Each VOP includes shape information and texture information to be separated. If the moving picture sequence includes a single VO, then the VOP is equated to a frame. There is no shape information included, in this case, and the texture information alone is then to be encoded/decoded.

The VOP includes alpha data representing the shape information and texture data representing the texture information, as illustrated in FIG. 30. Each data are defined as an aggregate of blocks (alphablocks/macroblocks), and each block in the aggregate is composed of 16×16 samples. Each alphablock sample is represented in eight bits. A macroblock includes accompanied chrominance signals being associated with 16×16 sample luminance signals. VOP data are obtained from a moving picture sequence externally processed outside of an encoder.

FIG. 31 is a diagram showing the configuration of a VOP encoder according to the VM encoding system. The diagram includes original VOP data P1 to be inputted, an alphablock P2 representing the shape information of the VOP, a switch P3a for passing the shape information, if there is any, of the inputted original VOP data, a shape encoder P4 for compressing and encoding the alphablock, compressed alphablock data P5, a locally decoded alphablock P6, texture data (a macroblock) P7, a motion detector P8, a motion parameter P9, a motion compensator P10, a predicted picture candidate P11, a prediction mode selector P12, a prediction mode P13, a predicted picture P14, a prediction error signal P15, a texture encoder P16, texture encoding information P17, a locally decoded prediction error signal P18, a locally decoded macroblock P19, a sprite memory update unit P20, a VOP memory P21, a sprite memory P22, a variable-length encoder/multiplexer P23, a buffer P24, and an encoded bitstream P25.

FIG. 32 shows a flowchart outlining an operation of the encoder.

Referring to the encoder of FIG. 31, the original VOP data P1 are decomposed into the alphablocks P2 and the macroblocks P7 (Steps PS2 and PS3). The alphablocks P2 and the macroblocks P7 are transferred to the shape encoder P4 and the motion detector P8, respectively. The shape encoder P4 is a processing block for data compression of the alphablock P2 (step PS4), the process of which is not discussed here further in detail because the compression method of shape information is not particularly relevant to the present invention.

The shape encoder P4 outputs the compressed alphablock data P5 which is transferred to the variable-length encoder/multiplexer P23, and the locally decoded alpha data P6 which is transferred sequentially to the motion detector P8, the motion compensator P10, the prediction mode selector P12, and the texture encoder P16.

The motion detector P8, upon reception of the macroblock P7, detects a local-motion vector on a macroblock basis using reference picture data stored in the VOP memory P21 and the locally decoded alphablock P6 (step PS5). Here, the motion vector is one example of a motion parameter. The VOP memory P21 stores the locally decoded picture of a previously encoded VOP. The content of the VOP memory P21 is sequentially updated with the locally decoded picture of a macroblock whenever the macroblock is encoded. In addition, the motion detector P8 detects a global warping parameter, upon reception of the full texture data of the original VOP, by using reference picture data stored in the sprite memory P22 and locally decoded alpha data. The sprite memory P22 will be discussed later in detail.

The motion compensator P10 generates the predicted picture candidate P11 by using the motion parameter P9, which is detected in the motion detector P8, and the locally decoded alphablock P6 (step PS6). Then, the prediction mode selector P12 determines the final of the predicted picture P14 and corresponding prediction mode P13 of the macroblock by using a prediction error signal power and an original signal power (step PS7). In addition, the prediction mode selector P12 judges the coding type of the data either intra-frame coding or inter-frame coding.

The texture encoder P16 processes the prediction error signal P15 or the original macroblock through Discrete Cosine Transformation (DCT) and quantization to obtain a quantized DCT coefficient based upon the prediction mode P13. An obtained quantized DCT coefficient is transferred, directly or after prediction, to the variable-length encoder/multiplexer P23 to be encoded (steps PS8 and PS9). The variable-length encoder/multiplexer P23 converts the received data into a bitstream and multiplexes the data based upon predetermined syntaxes and variable-length codes (step PS10). The quantized DCT coefficient is subject to dequantization and inverse DCT to obtain the locally decoded prediction error signal P18, which is added to the predicted picture P14, and the locally decoded macroblock P19 (step PS11) is obtained. The locally decoded macroblock P19 is written into the VOP memory P21 and the sprite memory P22 to be used for a later VOP prediction (step PS12).

Dominant portions of prediction including a prediction method, a motion compensation, and the update control of the sprite memory P22 and the VOP memory P21 will be discussed below in detail.

Normally, four different types of VOP encoding shown in FIG. 33 are processed in the VM. Each encoding type is associated with a prediction type or method marked by a circle on a macroblock basis. With an I-VOP, intra-frame coding is used singly involving no prediction. With a P-VOP, past VOP data can be used for prediction. With a B-VOP, both past and future VOP data can be used for prediction.

All the aforementioned prediction types are motion vector based. On the other hand, with a Sprite-VOP, a sprite memory can be used for prediction. The sprite is a picture space generated through a step-by-step mixing process of VOPs based upon a warping parameter set

• Provisional Patent
• Utility Patent

• What Is a Patent?
• What Is a Trademark or Servicemark?
• What Is a Copyright?
• Patent Laws