Introduction to MPEG-4
MPEG-4 introduces a number of new coding tools as shown in Figure 1.13. In MPEG-1 and MPEG-2 the motion compensation is based on regular fixed-size areas of image known as macroblocks. Whilst this works well at the designed bit rates, there will always be some inefficiency due to real moving objects failing to align with macroblock boundaries. This will increase the residual bit rate. In MPEG-4, moving objects can be coded as arbitrary shapes.
Figure 1.14 shows that a background can be coded quite independently from objects in front of it. Object motion can then be described with vectors and much-reduced residual data. According to the Profile, objects may be twodimensional, three dimensional and opaque or translucent. The decoder must contain effectively a layering vision mixer which is capable of prioritizing image data as a function of how close it is to the viewer. The picture coding of MPEG-4 is known as texture coding and is more advanced than the MPEG-2 equivalent, using more lossless predictive coding for pixel values, coefficients and vectors.
In contrast, MPEG-4 may move the rendering process to the decoder, reducing the bit rate needed with the penalty
of increased decoder complexity.
In addition to motion compensation, MPEG-4 can describe how an object changes its perspective as it moves using a technique called mesh coding. By warping another image, the prediction of the present image is improved.
MPEG-4 also introduces coding for still images using DCT or wavelets.
Although MPEG-2 supported some scaleability, MPEG-4 also takes this further. In addition to spatial and noise scaleability, MPEG-4 also allows temporal scaleability where a base level bitstream having a certain frame rate may be augmented by an additional enhancement bitstream to produce a decoder output at a higher frame rate.
This is important as it allows a way forward from the marginal frame rates of today’s film and television formats whilst remaining backwards compatible with traditional equipment. The comprehensive scaleability of MPEG-4 is equally important in networks where it allows the user the best picture possible for the available bit rate.
MPEG-4 also introduces standards for face and body animation. Specialized vectors allow a still picture of a face and optionally a body to be animated to allow expressions and gestures to accompany speech at very low bit rates.
In some senses MPEG-4 has gone upstream of the video signal which forms the input to MPEG-1 and MPEG-2 coders to analyse ways in which the video signal was rendered. Figure 1.14(a) shows that in a system using MPEG-1 and MPEG-2, all rendering and production steps take place before the encoder. Figure 1.14(b) shows that in MPEG-4, some of these steps can take place in the decoder. The advantage is that fewer data need to be transmitted. Some of these data will be rendering instructions which can be very efficient and result in a high compression factor. As a significant part of the rendering takes place in the decoder, computer graphics generators can be designed directly to output an MPEG-4 bitstream. In interactive systems such as simulators and video games, inputs from the user can move objects around the screen. The disadvantage is increased decoder complexity, but as the economics of digital processing continues to advance this is hardly a serious concern.
As might be expected, the huge range of coding tools in MPEG-4 is excessive for many applications. As with MPEG-2 this has been dealt with using Profiles and Levels. Figure 1.15 shows the range of Visual
.bmp)
MPEG-4 introduces a number of new coding tools as shown in Figure 1.13. In MPEG-1 and MPEG-2 the motion compensation is based on regular fixed-size areas of image known as macroblocks. Whilst this works well at the designed bit rates, there will always be some inefficiency due to real moving objects failing to align with macroblock boundaries. This will increase the residual bit rate. In MPEG-4, moving objects can be coded as arbitrary shapes.
Figure 1.14 shows that a background can be coded quite independently from objects in front of it. Object motion can then be described with vectors and much-reduced residual data. According to the Profile, objects may be twodimensional, three dimensional and opaque or translucent. The decoder must contain effectively a layering vision mixer which is capable of prioritizing image data as a function of how close it is to the viewer. The picture coding of MPEG-4 is known as texture coding and is more advanced than the MPEG-2 equivalent, using more lossless predictive coding for pixel values, coefficients and vectors.
In contrast, MPEG-4 may move the rendering process to the decoder, reducing the bit rate needed with the penaltyof increased decoder complexity.
In addition to motion compensation, MPEG-4 can describe how an object changes its perspective as it moves using a technique called mesh coding. By warping another image, the prediction of the present image is improved.
MPEG-4 also introduces coding for still images using DCT or wavelets.
Although MPEG-2 supported some scaleability, MPEG-4 also takes this further. In addition to spatial and noise scaleability, MPEG-4 also allows temporal scaleability where a base level bitstream having a certain frame rate may be augmented by an additional enhancement bitstream to produce a decoder output at a higher frame rate.
This is important as it allows a way forward from the marginal frame rates of today’s film and television formats whilst remaining backwards compatible with traditional equipment. The comprehensive scaleability of MPEG-4 is equally important in networks where it allows the user the best picture possible for the available bit rate.
MPEG-4 also introduces standards for face and body animation. Specialized vectors allow a still picture of a face and optionally a body to be animated to allow expressions and gestures to accompany speech at very low bit rates.
In some senses MPEG-4 has gone upstream of the video signal which forms the input to MPEG-1 and MPEG-2 coders to analyse ways in which the video signal was rendered. Figure 1.14(a) shows that in a system using MPEG-1 and MPEG-2, all rendering and production steps take place before the encoder. Figure 1.14(b) shows that in MPEG-4, some of these steps can take place in the decoder. The advantage is that fewer data need to be transmitted. Some of these data will be rendering instructions which can be very efficient and result in a high compression factor. As a significant part of the rendering takes place in the decoder, computer graphics generators can be designed directly to output an MPEG-4 bitstream. In interactive systems such as simulators and video games, inputs from the user can move objects around the screen. The disadvantage is increased decoder complexity, but as the economics of digital processing continues to advance this is hardly a serious concern.
As might be expected, the huge range of coding tools in MPEG-4 is excessive for many applications. As with MPEG-2 this has been dealt with using Profiles and Levels. Figure 1.15 shows the range of Visual
0 comments:
Post a Comment