Method and apparatus to prioritize video information during coding and decoding

[0002] This application is a rule 1.53(b) Continuation Application of and claims priority to U.S. patent application Ser. No. 08/986,118 filed Dec. 5, 1997 now abandoned, which claims benefit of Provisional Appln No. 60/038,115 filed Feb. 20, 1997, which is herein incorporated by reference.

[0003] The present invention relates to video coding and, more particularly, to the use of video objects in combination with keyregions to improve coding efficiency and image quality.

[0004] The advent of video objects and video object planes (VOPs) in video coding permits significant coding savings by selectively apportioning bits among portions of the frame that require a relatively large amount of bits and other portions that require a relatively small number of bits. VOPs also permit additional functionality such as object manipulation.

[0005] As an example, FIG. 1(a) illustrates a frame for coding that includes a head and shoulders of a narrator, a logo suspended within the frame and a background. The logo may be static, having no motion and no animation. In such a case, bit savings may be realized by coding the logo only once. For display, the coded logo could be decoded and displayed continuously from the single coded representation. Similarly, it may be desirable to code the background at a low refresh rate to save bits and yet create an illusion of movement in the reconstructed image. Bit savings realized by coding the logo and background at lower rates may permit coding of the narrator at a higher rate, where the perceptual significance of the image may reside. VOPs are suited to such applications. FIGS. 1(b)-1(d) illustrate the frame of FIG. 1(a) broken into three VOPs. By convention, a background generally is assigned VOPØ. The narrator and logo may be assigned VOP1 and VOP2 respectively. Within each VOP, all image data is coded and decoded identically.

[0006] Not all data within a VOP merits identical treatment. For example, certain regions of a VOP may require animation, whereas others are relatively static. Consider the example of the narrator. The perceptually significant areas of VOP1 center around the facial features of the figure. The narrator's dress and hair may not require animation to the same extent that the facial features do. Accordingly, there is a need in the art for a coding system that emphasizes certain areas of a VOP over others.

[0007] Further, regions of a VOP may possess similar characteristics. For example, some image data within the VOP may exhibit the same motion vector or may be quantized according to the same quantization parameters. Certain regions of a VOP may require a greater resilience against channel errors. Coding efficiencies may be obtained by coding the similar attributes only once for the region. Such efficiencies are lost unless coding systems provide a means for coding common attributes of the region differently from other regions in a VOP that do not share the common attributes.

[0008] Finally, it may be preferable to embed functionalities in certain regions of a VOP. For example, images may be superimposed upon regions of a VOP. Consider an example where it is desirable to impose a logo image onto the dress of the narrator in VOPI and permit a viewer to selectively enable or disable display of the logo. Accordingly, there is a need in the art to associate functionalities with certain regions of a VOP.

[0009] The present invention alleviates the aforementioned needs in the art to a great extent by providing keyregions for VOPs. Keyregions exhibit one or more of the following properties:

[0010] they are optional,

[0011] they consist of a sequence of macroblocks,

[0012] they are two-dimensional but do not have to be rectangular,

[0013] a VOP may be portioned into keyregions, but not every macroblock of a VOP must belong to a keyregion,

[0014] a macroblock that is not a member of a keyregion may be a member of a background keyregion,

[0015] keyregions begin and end in the same VOP,

[0016] a macroblock that belongs to a keyregion belongs to only one keyregion, and

[0017] macroblocks of a keyregion share at least one common attribute.

[0018] The keyregion is defined in the coded video information by a keyregion header that identifies the location and width of the keyregion. Data of the macroblocks in the keyregion is decoded by a decoder and placed sequentially within the keyregion until the decoder receives another keyregion header.

[0019] FIG. 1(a) illustrates a video frame to be coded according to the present invention; FIGS. 1(b)-1(d) represent video objects from the frame of FIG. 1(a) to be coded according to the present invention.

[0020] FIG. 2 is a block diagram of the present invention.

[0021] FIG. 3 represents the structure of a keyregion header generated according to the present invention.

[0022] FIG. 4 illustrates a video object coded by keyregion according to the present invention.

[0023] FIG. 5 illustrates the operation of a decoder operating in accordance with the present invention.

[0024] The present invention provides keyregions to code areas of VOPs at lower bit rates and with improved image quality. Keyregions are a collection of macroblocks within a VOP that are related according to one or more attributes. For example, the macroblocks within a keyregion may have been quantized according to the same quantization parameter, they may exhibit the same motion vector and/or they may possess the same priority. Typically, however, the macroblocks do not merit coding as a separate VOP as the bit costs associated with VOP coding would result in coding inefficiencies. Based on similarities among the macroblocks, coding efficiencies are obtained by organizing the macroblocks into keyregions and coding common information only once.

[0025] According to the present invention, as shown in FIG. 2, an encoder 100 receives a video signal representative of a frame or frames to be coded. The video signal is sampled and organized into macroblocks which are spatial areas of each frame. The encoder 100 codes the macroblocks and outputs an encoded bitstream to a channel 150. The bitstream may identify some macroblocks as having been organized and coded as VOPs. The channel 150 may be a radio channel, a computer network or some storage media such as a memory or a magnetic or optical disk. A decoder 200 retrieves the bitstream from the channel 150 and reconstructs a video signal therefrom for display.

[0026] The encoder 100 defines a VOP in the bitstream by generating a VOP header. VOP headers define the position, shape and size of the VOP. As is known, the shape of a VOP may be defined to a pixel or two pixel level. After decoding a VOP header, the decoder 200 knows which macroblocks or portions of macroblocks are members of the VOP and which are not. When implemented with the present invention, the VOP header contains a keyregion enable signal indicating whether the VOP contains one or more keyregions. The keyregion enable signal may be as short as a single bit in a predetermined position of the VOP header.

[0027] In the bitstream, a keyregion is defined by an overhead signal, called a “keyregion header,” followed by data for macroblocks of the keyregion. FIG. 3 illustrates the data structure of the keyregion header 300. To indicate the occurrence of a keyregion, the encoder 100 generates a resync marker 310, a code that possesses a unique predetermined bit sequence. The resync marker sequence cannot occur naturally in the VOP. The encoder 100 also generates a macroblock number signal 320 identifying a macroblock that is the origin of the keyregion. For example, the macroblock number 320 may define an address of the upper left hand corner of the keyregion. The macroblock number 320 is a code whose length is determined by the size of the VOP in which the keyregion resides. Also, the encoder 100 generates a keyregion width signal 330 defining the keyregion's width in terms of macroblocks. Again, keyregion width field 330 possesses a length that is determined by the size of the VOP in which the keyregion resides.

[0028] The macroblock number and width fields 320 and 330 define a bounding box that circumscribes the keyregion. For example, to code keyregion K1 within VOP1 (shown in FIG. 4), the macroblock number field identifies macroblock MB1 as the keyregion's origin. The width field 330 defines the keyregion to be four macroblocks wide. These fields define a bounding box B1 bounded by lines L1, L2 and L3. The bottom edge of the bounding box B1 is not defined by the macroblock number and width fields. By default, the keyregion is defined to occupy the entire area of the bounding box B1 that falls within the area of VOP1. Thus, the default keyregion includes macroblocks MB3-4, MB6-16, etc. In the VOP header, macroblock MB1, MB2 and MB5 would have been defined to be excluded from VOP 1. If the keyregion takes an irregular shape, as keyregion K1 does in FIG. 4, the shape is defined by a shape refine field 340.

[0029] The shape refine field 340 follows the width field 330 in the keyregion header 300. It contains a shape refine flag 342. The shape refine flag 342 is a one bit code that, when enabled, indicates that the keyregion takes an arbitrary shape. If the flag 342 is enabled, it is followed by a shape code 344 that identifies which of the macroblocks contained within a bounding box of the keyregion is found within the keyregion. The shape code 344 provides a bit associated with each macroblock contained in the bounding box provided the macroblock falls within the area of the VOP. The status of the bit determines whether the associated macroblock is included in the keyregion. If the flag 342 is disabled, the shape code 342 is omitted from the shape refine field 340.

[0030] Consider again, VOP1 in FIG. 4. As noted, the macroblock number 320 and width 330 fields define the bounding block to every macroblock of VOP1 that falls within a column that begins at macroblock MB1, extends laterally four macroblocks from macroblock MB1 and extends vertically to the bottom of VOP1. However, keyregion K1 is irregular. It includes only macroblocks MB3, MB4, MB6-11 and MB13-15. To define the irregular shape of the keyregion, the shape code 342 will be a thirteen bit code that identifies which macroblocks are part of the irregular keyregion. The following table demonstrates how the shape code 344 defines the membership of each macroblock in keyregion K1:

[0031] Again, data for macroblocks MB1, MB2 and MB5 s not provided in the shape code 344 because were defined as not being members of VOP1.

[0032] If a shape code 344 is included in the keyregion header 300, the shape code 344 identifies how many macroblock are contained in the keyregion.

[0033] The keyregion header 300 also identifies data that is common throughout the keyregion. For example, a quantization parameter field 350, a motion compensation field 360 and a priority field 370 may be provided for any keyregion. Preferably, each may be identified in the keyregion header 300 by a one bit flag that, if enabled, is followed by a code representing the attribute value. The keyregion may possess more than one common attribute. The decoder 200 uses the common attribute information to decode the macroblocks that follow the keyregion header 300. The keyregion header 300 is followed by a variable length sequence of macroblock data (not shown) representing image data of the macroblocks within the keyregion. For keyregion macroblocks that overlap the edge of the VOP, the decoder interprets encoded data to represent only the portion of the macroblock that falls within the VOP according to conventional encoding.

[0034] At the time of this writing, the MPEG-4 video standard is being drafted. The keyregion coding scheme of the present invention has been proposed for integration into the MPEG-4 video standard. Under such proposal, the resync marker 310 is defined as bit sequence of sixteen zeros and a one (“0000 0000 0000 0000 1”). The macroblock number 320 code is a 1-12 bit code representing an address of the top left corner of the bounding box. The code length is determined by the following formula: $\text{Length}=\frac{\mathrm{VOP}\mathrm{width}}{16}\times \frac{\mathrm{VOP}\mathrm{height}}{16}.$

[0035] The width field 330 is a 1-7 bit code representing the width of the keyregion in macroblock units. Again, the length of the width field depends upon the width of the VOP. The shape refine field 340 is a one bit code. The quantization parameter value, the priority value and the motion vector values are each two bit codes.

[0036] FIG. 5 illustrates a method of operation of the decoder 200. The decoder 200 detects the keyregion when it detects the resync marker (Step 1010). The decoder 200 decodes the keyregion header 300 to construct the keyregion. The decoder detects the macroblock number 320 and width 330 fields define the bounding box B1 circumscribing the keyregion K1 (Steps 1020 and 1030). By default, the decoder 200 defines the keyregion to include every macroblock that falls within the union of the VOP and the bounding box B1 (Step 1040). However, the decoder 200 receives the shape refine field (Step 1050). If the shape refine flag 342 is enabled (Step 1060), the decoder 200 decodes the shape code data 344 (Step 1070) to identify macroblocks from the bounding box B1 that are excluded from the keyregion (Step 1080).

[0037] The decoder 200 receives and decodes the common attribute data (Step 1090). Using the attribute data, the decoder 200 receives and decodes macroblock data and places each macroblock sequentially in position according to a raster scan direction (left to right, then down) over the keyregion (Steps 1100 and 1110). The decoder 200 does not place decoded data in any position that is not included within the keyregion.

[0038] After receiving and decoding the keyregion header 300, the decoder 200 receives all successive data as macroblock data until it is interrupted. For example, following the macroblock data of the keyregion, the bitstream may include another resync marker indicating the beginning of another keyregion. Alternatively, the bitstream may include a succeeding VOP header or another data pattern indicating the occurrence of another type of data. When the decoder 200 detects such a data pattern in the bitstream, it ceases to decode data as macroblock data associated with the keyregion.

[0039] The present invention provides a system for encoding and decoding keyregions in video object planes. The keyregions realizes efficient coding of VOP data when a portion of the data share common attributes that are not distributed throughout the entire VOP. For example, when a specific region of a VOP requires coding at a higher resolution than the remainder of the VOP, a single quantization parameter may be set for the region using the keyregion of the present invention. Coding of the high resolution image segment occurs using a greater number of bits than the remainder of the VOP. In this way, bits are conserved in coding the VOP remainder. Also, motion information or priority information may be coded for a keyregion, yielding coding efficiencies that would not be realized if the attribute data were either distributed to the entire VOP or established on a macroblock by macroblock basis.