KR101166732B1 - Video coding mode selection using estimated coding costs - Google Patents

Abstract

This disclosure describes techniques related to video coding mode selection using estimated coding costs. In order to provide high compression efficiency, for example, the encoding device may attempt to select a coding mode for coding a block of pixels that codes data of blocks with high efficiency. In turn, this coding device may perform coding mode selection based on an estimate of the coding cost for at least some of the possible modes. According to the techniques described herein, the encoding device estimates the coding cost for different modes without actually coding the blocks. In fact, in some aspects, the encoding module device may estimate the coding cost for the modes without quantizing the block's data for each mode. In this way, the coding cost estimation technique of the present disclosure reduces the computationally intensive computational amount required to perform effective mode selection.

Coding mode selection, coding cost estimation, transform coefficients, quantization

Description

VIDEO CODING MODE SELECTION USING ESTIMATED CODING COSTS}

Technical Field

TECHNICAL FIELD This disclosure relates to video coding, and more particularly, to estimating a coding cost for coding a video sequence.

Background technology

Digital video capabilities include digital televisions, digital direct broadcast systems, wireless communication devices, personal digital assistants (PDAs), laptop computers, desk computers, video game consoles, digital cameras, digital recording devices, cellular or satellite cordless phones, and the like. Can be integrated into a wide range of devices. Digital video devices can provide significant improvements over conventional analog video systems in the processing and transmission of video sequences.

Different video coding standards have been established for coding digital video sequences. The Video Experts Group (MPEG) has developed a number of standards, including, for example, MPEG-1, MPEG-2 and MPEG-4. Other examples include the International Telecommunication Union (ITU) -T H.263 standard, and the ITU-T H.264 standard and its counterparts, ISO / IEC MPEG-4, Part 10, or Advanced Video Coding (AVC). . This video coding standard supports improving the communication efficiency of video sequences by coding the data in a compressed manner.

Many current technologies use block based coding. In block-based coding, a frame of a multimedia sequence is divided into individual blocks of pixels, and blocks of pixels are coded based on differences from other blocks that may be located in the same frame or in different frames. Blocks of some pixels, sometimes referred to as "macroblocks", include grouping of subblocks of pixels. By way of example, a 16x16 macroblock may include four 8x8 subblocks. Subblocks may be coded separately. For example, the H.264 standard codes blocks in a variety of different sizes, for example 16 × 16, 16 × 8, 8 × 16, 8 × 8, 4 × 4, 8 × 4, and 4 × 8. To do it. Furthermore, further subblocks of any size may be included in macroblocks, for example 2x16, 16x2, 2x2, 4x16, and 8x2.

summary

In certain aspects of the present disclosure, a method of processing digital video data includes identifying one or more transform coefficients for residual data of a block of pixels that will remain non-zero when quantized, based at least on the identified transform coefficients. Estimating the number of bits associated with the coding of the residual data, and estimating the coding cost for coding the block of pixels based at least on the estimated number of bits associated with the coding of the residual data.

In a particular aspect, an apparatus for processing digital video data includes a transform module that generates transform coefficients for residual data of a block of pixels, identifies one or more transform coefficients that will remain non-zero when quantized and at least identified transform coefficients A bit estimation module for estimating the number of bits associated with the coding of the residual data based on the data, and a control for estimating the coding cost for coding the block of pixels based at least on the estimated number of bits associated with the coding of the residual data. Contains modules

In a particular aspect, an apparatus for processing digital video data includes means for identifying one or more transform coefficients for residual data of a block of pixels that will remain non-zero when quantized, the residual data based at least on the identified transform coefficients Means for estimating the number of bits associated with the coding of, and means for estimating the coding cost for coding the block of pixels based at least on the estimated number of bits associated with the coding of the residual data.

In a particular aspect, a computer program product for processing digital video data includes a computer readable medium having instructions. These instructions are code that identifies one or more transform coefficients for the residual data of the block of pixels that will remain non-zero when quantized, and estimate the number of bits associated with coding of the residual data based at least on the identified transform coefficients. Code for estimating a coding cost for coding a block of pixels based on at least an estimated number of bits associated with coding of residual data.

The details of one or more examples are set forth in the accompanying drawings and the description below. Other features, objects, and advantages will be apparent from the description and drawings, and from the claims.

Brief description of the drawings

1 is a block diagram illustrating a video coding system utilizing the coding cost estimation technique described herein.

2 is a block diagram illustrating an example encoding module in more detail.

3 is a block diagram illustrating another exemplary encoding module in more detail.

4 is a flowchart illustrating an exemplary operation of an encoding module to select an encoding mode based on an estimated coding cost.

5 is a flowchart illustrating an exemplary operation of an encoding module to estimate the number of bits associated with coding of residual data of a block without encoding or quantizing the residual data.

6 is a flowchart illustrating an exemplary operation of an encoding module to estimate the number of bits associated with coding of residual data of a block without encoding residual data.

details

This disclosure describes techniques related to video coding mode selection using estimated coding costs. In order to provide high compression efficiency, for example, the encoding device may attempt to select a coding mode for coding a block of pixels that codes the data of the blocks at high efficiency. In turn, this coding device may perform coding mode selection based on an estimate of at least coding cost for at least some of the possible modes. According to the techniques described herein, the encoding device estimates the coding cost for different modes without actually coding the blocks. In fact, in some aspects, the encoding module device may estimate the coding cost for the modes without quantizing the block's data for each mode. In this way, the coding cost estimation technique of the present disclosure reduces the computationally intensive computational amount required to perform effective mode selection.

1 is a block diagram illustrating a multimedia coding system 10 that employs a coding cost estimation technique as described herein. Coding system 10 includes an encoding device 12 and a decoding device 14 connected by a transport channel 16. Encoding device 12 encodes, decodes and possibly decodes the encoded sequence via transport channel 16 for presentation to a user of decoding device 14. Transmit to device 14. Transport channel 16 may comprise any wired or wireless medium, or a combination thereof.

Encoding device 12 may form part of a broadcast network component used to broadcast one or more channels of multimedia data. By way of example, encoding device 12 may form part of a wireless base station, server, or any infrastructure node used to broadcast one or more channels of encoded multimedia data to a wireless device. In this case, encoding device 12 may transmit the encoded data to a plurality of wireless devices, such as decoding device 14. However, only one decoding device 14 is shown in FIG. 1 for the sake of simplicity. Alternatively, encoding device 12 may include a handset that transmits locally captured video for video telephony or other similar application.

Decoding device 14 may include a user device that receives encoded multimedia data sent by encoding device 12 and decodes the multimedia data to present to the user. By way of example, decoding device 14 may be a digital music, video device, such as a digital television, wireless communication device, gaming device, portable digital assistant (PDA), laptop computer or desktop computer, products sold under the trade name “iPod”. Or as part of a wireless telephone such as a cellular, satellite, or terrestrial based telephone, or other wireless mobile terminal with video and / or audio streaming, video telephony, or both. Decoding device 14 may be associated with a mobile or fixed device. In a broadcast application, encoding device 12 may transmit the encoded video and / or audio to multiple decoding devices 14 associated with multiple users.

In some aspects, for bidirectional communication applications, the multimedia coding system 10 may include a Session Initiation Protocol (SIP), International Telecommunication Union Standards Sector (ITU-T) H.323 Standard, ITU-T H.324 Standard, or other. Depending on the standard, it may also support video telephony or video streaming. For one-way or two-way communication, encoding device 12 corresponds to Video Expert Group (MPEG) -2, MPEG-4, ITU-T H.263, or MPEG-4, Part 10, Advanced Video Coding (AVC). It is also possible to generate encoded multimedia data according to video compression standards such as ITU-T H.264. Although not shown in FIG. 1, encoding device 12 and decoding device 14 may be integrated into an encoder and a decoder, respectively, and may be appropriate multiplexers to handle the encoding of both audio and video into a common data sequence or a separate data sequence. A demultiplexer (MUX-DEMUX) module, or other hardware, firmware, or software. If applicable, MUX-DEMUX modules may follow another protocol, such as the ITU H.223 Multiplexer Protocol, or User Datagram Protocol (UDP).

In a particular aspect, the present disclosure relates to enhanced H.264 video coding that delivers real-time multimedia services in terrestrial mobile multimedia multicast (TM3) systems using the Forward Link Only (FLO) air interface specification. Consider the "Forward Link Only Air Interface Specification for Terrestrial Mobile Multimedia Multicast" ("FLO Specification") published as standard TIA-1099. However, the coding cost estimation technique described in this disclosure is not limited to any particular type of broadcast, multicast, unicast, or point-to-point system.

As shown in FIG. 1, encoding device 12 includes an encoding module 18 and a transmitter 20. Encoding module 18, in the case of video encoding, receives one or more input multimedia sequences that may include one or more frames of data and optionally encodes the frames of the received multimedia sequence. Encoding module 18 receives an input multimedia sequence from one or more sources (not shown in FIG. 1). In some aspects, encoding module 18 may receive an input multimedia sequence from one or more video content providers, eg, satellites. As another example, encoding module 18 may receive a multimedia sequence from an image capture device (not shown in FIG. 1) integrated in or coupled to encoding device 12. Alternatively, encoding module 18 may receive a multimedia sequence from a memory or archive (not shown in FIG. 1) within or coupled to encoding device 12. The multimedia sequence may comprise a live real-time or near real-time video, audio, or video and audio sequence that is coded and transmitted as a broadcast or on-demand, or coded to be broadcast or on-demand And may include pre-recorded and stored video, audio, or video and audio sequences that are transmitted. In some aspects, at least some of the multimedia sequences may be computer generated as in the case of gaming.

In some cases, encoding module 18 encodes the plurality of coded frames and transmits via transmitter 20 to decoding device 14. Encoding module 18 may encode the frames of the input multimedia sequence as intra coded frames, inter coded frames, or a combination of both. Frames encoded using intra coding techniques are coded independently of other frames and are often referred to as intra (“I”) frames. Frames encoded using inter coding techniques are coded with respect to one or more other frames. Inter coded frames may include one or more prediction (“P”) frames, bidirectional (“B”) frames, or a combination thereof. P frames are encoded in relation to at least one temporal previous frame, while B frames are encoded in relation to at least one temporal future frame. In some cases, a B frame may be encoded with respect to at least one temporal future frame and at least one temporal previous frame.

Encoding module 18 may be further configured to divide the frame into a plurality of blocks and to encode each of the blocks individually. By way of example, encoding module 18 may split the frame into a plurality of 16 × 16 blocks. Often, some blocks called "macroblocks" comprise groupings of subdivided blocks (herein referred to as "sub-blocks"). By way of example, a 16x16 macroblock may include four 8x8 subblocks, or other subdivision blocks. For example, the H.264 standard encodes blocks having various different sizes, for example 16 × 16, 16 × 8, 8 × 16, 8 × 8, 4 × 4, 8 × 4, and 4 × 8. Allow. In addition, when expanded, a subblock of any size may be included in a macro block, for example, 2x16, 16x2, 2x2, 4x16, 8x2, or the like. Thus, encoding module 18 may be configured to divide the frame into several blocks and encode each of the blocks of pixels as an intra coded block or an inter coded block, wherein the intra coded block or inter coded block is Each may also be referred to generally as a block.

Encoding module 18 may support multiple coding modes. Each of the modes may correspond to a different combination of block size and coding technique. For the H.264 standard, for example, there are seven inter modes and thirteen intra modes. The seven variable block size inter modes include SKIP mode, 16x16 mode, 16x8 mode, 8x16 mode, 8x8 mode, 8x4 mode, 4x8 mode, and 4x4 mode. The thirteen intra modes include an INTRA 4x4 mode with nine possible interpolation directions and an INTRA 16x16 mode with four possible interpolation directions.

In order to provide high compression efficiency, in accordance with various aspects of the present disclosure, encoding module 18 attempts to select a mode for coding the data of the block at high efficiency. In turn, encoding module 18 estimates, for each of the blocks, the coding cost for at least some of the modes. Encoding module 18 estimates the coding cost as a function of rate and distortion. According to the techniques described herein, encoding module 18 estimates the coding cost for the modes without actually coding the block to determine the rate and distortion metric. In this way, encoding module 18 may select one of the modes based at least on the coding cost without performing computationally complex coding of the data of the block for each mode. Conventional mode selection requires actual coding of the data using each of the modes to determine the mode to select. As such, the present technology saves time and computational resources by selecting a mode based on coding cost without actually coding the data for each of the modes. In fact, in some aspects, encoding module 18 may estimate the coding cost for the modes without quantizing the block's data for each mode. In this way, the coding cost estimation technique of the present disclosure reduces the computationally intensive computation required to perform effective mode selection.

Encoding device 12 applies the selected mode to code a block of frames and transmits the coded frame of data via transmitter 20. Transmitter 20 may include suitable modem and driver circuit software and / or firmware to transmit encoded multimedia over transport channel 16. In a wireless application, transmitter 26 includes RF circuitry for transmitting wireless data that carries encoded multimedia data.

Decoding device 14 includes a receiver 22 and a decoding module 24. Decoding device 14 receives encoded data from encoded device 12 via receiver 22. Similar to transmitter 20, receiver 22 may include suitable modem and driver circuit software and / or firmware to receive encoded multimedia over transmission channel 16, and transmit encoded multimedia data in a wireless application. It may also include RF circuitry to receive the wireless data to communicate. Decoding module 24 decodes a coded frame of data received via receiver 22. Decoding device 14 may also be integrated into decoding device 14 to a user via a display (not shown), which may be provided as a separate device coupled to decoding device 14 via a wireless connection. It may represent a decoded frame of data.

In some examples, encoding device 12 and decoding device 14 each include mutual transmission and reception circuitry, each transmitting and receiving devices for encoded multimedia and other information transmitted over transmission channel 16. It can also serve as both devices. In this case, both encoding device 12 and decoding device 14 may transmit and receive a multimedia sequence, resulting in participating in bidirectional communication. In other words, the described components of coding system 10 may be integrated as part of an encoder / decoder (CODEC).

The components in encoding device 12 and decoding device 14 are examples of components applicable for implementing the techniques described herein. However, if desired, encoding device 12 and decoding device 14 may include many other components. For example, encoding device 12 may include a plurality of encoding modules, each of which receives one or more sequences of multimedia data and encodes each sequence of multimedia data in accordance with the techniques described herein. In this case, encoding device 12 may further include at least one multiplexer that combines segments of data for transmission. In addition, encoding device 12 and decoding device 14, depending on the application, include appropriate modulation, demodulation, frequency conversion, for transmission and reception of encoded video, including radio frequency (RF) radio components and antennas. Filtering, and amplifier components. However, for ease of explanation, these components are not shown in FIG. 1.

2 is a block diagram illustrating an example encoding module 30 in more detail. Encoding module 30 may represent, for example, encoding module 18 of encoding device 12 of FIG. 1. As shown in FIG. 2, encoding module 30 includes a control module 32 that receives an input frame of multimedia data of one or more multimedia sequences from one or more sources and processes the frame of the received multimedia sequence. . In particular, the control module 32 analyzes the incoming frames of the multimedia sequence and determines whether to encode or skip the incoming frames based on the analysis of those frames. In some aspects, encoding device 12 may encode the information included in the multimedia sequence at a reduced frame rate using frame skipping to conserve bandwidth over transport channel 16.

Moreover, for incoming frames to be encoded, control module 32 may also be configured to determine whether to encode the frame as an I frame, a P frame or a B frame. Control module 32 may determine to encode the incoming frame into an I frame at the beginning of the multimedia sequence, at a scene change in the sequence, for use as a channel switch frame, or for use as an intra refresh frame. Otherwise, control module 32 encodes the frame as an inter coded frame (ie, a P frame or a B frame) to reduce the amount of bandwidth associated with coding of the frame.

Control module 32 may also be configured to divide the frame into a plurality of blocks and, for each of the blocks, select a coding mode, such as one of the H.264 coding modes described above. As will be described in more detail below, encoding module 30 may estimate a coding cost for at least some of the modes to help select the most efficient one of the coding modes. After selecting the coding mode to use to code one of the blocks, encoding module 30 generates residual data for the block. For the block selected to be intra coded, spatial prediction module 34 generates residual data for the block. Spatial prediction module 34 may, for example, generate a predicted version of the block through interpolation using interpolation directionality and one or more adjacent blocks corresponding to the selected intra coding mode. Spatial prediction module 34 may then calculate the difference between the block of the input frame and the predicted block. This difference is referred to as residual data or residual coefficient.

For the block selected to be inter coded, motion estimation module 36 and motion compensation module 38 generate residual data for the block. In particular, motion estimation module 36 identifies at least one reference frame and searches for a block within the reference frame that best matches a block within that input frame. Motion estimation module 36 calculates a motion vector to indicate an offset between the position of the block in the input frame and the position of the identified block in the reference frame. Motion compensation module 38 calculates the difference between the block of the input frame and the identified block in the reference frame, indicated by the motion vector. This difference is the residual data for the block.

Encoding module 30 also includes transform module 40, quantization module 46, and entropy encoder 48. Transform module 40 transforms the residual data of the block according to the transform function. In some aspects, transform module 40 applies an integer transform or discrete cosine transform (DCT), such as a 4x4 or 8x8 integer transform, to the residual data to generate transform coefficients for the residual data. Quantization module 46 quantizes the transform coefficients and provides the quantized transform coefficients to entropy encoder 48. Entropy encoder 48 encodes quantized transform coefficients using context-adaptive coding techniques, such as context-adaptive variable-length coding (CAVLC) or context-adaptive binary arithmetic coding (CABAC). . As will be described in more detail below, entropy encoder 48 applies the selected mode to code the data of the block.

Entropy encoder 48 may also encode additional data associated with the block. For example, in addition to the residual data, entropy encoder 48 may encode one or more motion vectors of the block, an identifier indicating the coding mode of the block, one or more reference frame indexes, quantization parameter (QP) information, slice information of the block, and the like. It may be. Entropy encoder 48 may receive this additional block data from other modules in encoding module 30. For example, motion vector information may be received from motion estimation module 36, while block mode information may be received from control module 32. In some aspects, entropy encoder 48 uses at least one of this additional information using universal variable length coding (VLC) or fixed length coding (FLC) techniques, such as exponential-Golomb coding (“Exp-Golomb”). You can also code some. Alternatively, entropy encoder 48 may encode some of the additional block data using the context adaptive coding techniques described above, ie, CABAC or CAVLC.

In order to assist the control module 32 in selecting the mode for the block, the control module 32 estimates the coding cost for at least some of the possible modes. In a particular aspect, control module 32 may estimate the cost of coding the block in each of the possible coding modes. This cost may, for example, be estimated with respect to the number of bits associated with coding a block in a given mode versus the amount of distortion generated in that mode. For the H.264 standard, for example, the control module 32 may include 22 different coding modes (inter coding mode and intra coding mode) for the block selected for inter coding and 13 for the block selected for intra coding. Coding costs may be estimated for two different coding modes. In another aspect, control module 32 may initially use all other selection techniques to reduce the set of possible modes, and then describe techniques of this disclosure to estimate coding costs for the remaining modes of the set. Can also be used. In other words, in some aspects, control module 32 narrows the number of mode possibilities before applying the cost estimation technique. Advantageously, encoding module 30 reduces the computational overhead associated with the coding decision by estimating the coding cost for the modes without actually coding the data of the blocks for the different modes. In fact, in the example shown in FIG. 2, encoding module 30 may estimate the coding cost without quantizing the data of the block for different modes. In this manner, the coding cost estimation technique of the present disclosure reduces the computationally intensive computational amount required to calculate the coding cost. In particular, there is no need to encode blocks using various coding modes to select one of the modes.

As will be described in more detail herein, the control module 32 estimates the coding cost of each analyzed mode according to equation (1).

J is the estimated coding cost, D is the distortion metric of the block, [lambda] mode is the Lagrangian multiplier of each mode, and R is the rate metric of the block. The distortion metric (D) may include, for example, sum of absolute difference (SAD), sum of square difference (SSD), sum of absolute transform difference (SATD), sum of square transform different (SSTD), and the like. . The rate metric R may be, for example, the number of bits associated with the coding of data in a given block. As mentioned above, different types of block data may be coded using different coding techniques. In this way, equation (1) may be rewritten in the form of equation (2).

Where R _context represents a rate metric for block data coded using a context adaptive coding technique and R _{non_context} represents a rate metric for block data coded using a non context adaptive coding technique. In the H.264 standard, for example, residual data may be coded using context adaptive coding such as CAVLC or CABAC. Other block data, such as motion vectors, block modes, may be coded using universal VLC or FLC techniques such as exponential-golom. In this case, equation (2) may be rewritten in the form of equation (3).

R _residual by using the context-adaptive coding technique for rate metric, for example, to code the residual data, indicates the number of bits associated with the encoding of the residual data, R _other different blocks by using the FLC or universal VLD Technology A rate metric for coding data, eg, the number of bits associated with coding of other block data.

In calculating the estimated coding cost J, encoding module 30 may use FLC or universal VLC to relatively easily determine the number of bits associated with the coding block data, ie, R _other . Encoding module 30 may, for example, use a code table to identify the number of bits associated with coding of block data using FLC or universal VLC. This code table may include, for example, a plurality of codewords and the number of bits associated with coding of the codeword. However, the determination of the number of bits associated with the coding of _residual data R _residual represents a much more difficult task due to the adaptive nature of context adaptive coding as a function of the context of the data. To determine the exact number of bits associated with the coding of residual data or which data is context adaptive coding, encoding module 30 transforms the residual data, quantizes the transformed residual data, and transform-quantizes the residual data. Must be encoded. However, in accordance with the techniques of this disclosure, bit estimation module 42 may estimate the number of bits associated with coding of residual data using a context adaptive coding technique without actually coding the residual data.

In the example shown in FIG. 2, bit estimation module 42 estimates the number of bits associated with the coding of the residual data using the transform coefficients for the residual data. As such, for each mode to be analyzed, encoding module 30 only needs to calculate transform coefficients for the residual data to estimate the number of bits associated with the coding of the residual data. Thus, encoding module 30 does not quantize the transform coefficients or encode the quantized transform coefficients for each of the modes so that the amount of time and computational resources required to determine the number of bits associated with the coding of the residual data. Decreases.

Bit estimation module 42 analyzes the transform coefficients output by transform module 40 to identify one or more transform coefficients that will remain non-zero after quantization. In particular, bit estimation module 42 compares each of the transform coefficients to a corresponding threshold. In some aspects, the corresponding threshold may be calculated as a function of QP of encoding module 30. Bit estimation module 42 identifies transform coefficients that are greater than or equal to their corresponding threshold, as transform coefficients that will remain non-zero after quantization.

Bit estimation module 42 estimates the number of bits associated with the coding of the residual data based at least on the transform coefficients identified as remaining non-zero after quantization. In particular, bit estimation module 42 may determine the number of non-zero transform coefficients that will remain in quantization. Bit estimation module 42 also sums at least some of the absolute values of the transform coefficients identified as remaining in quantization. The bit estimation module 42 then uses equation (4) to estimate the rate metric for the residual data, ie the number of bits associated with the coding of the residual data.

SATD is the sum of at least some of the absolute values of the non-zero transform coefficients predicted to remain in quantization, NZ _est is the estimated number of non-zero transform coefficients predicted to remain in quantization and a ₁ , a ₂ , And a ₃ are the coefficients. The coefficients a ₁ , a ₂ , and a ₃ may be calculated using, for example, a least squares estimate. The sum of the transform coefficients is the sum SATD of the absolute transform difference of the example of equation (4), but other difference coefficients such as SSTDs may be used.

An example calculation of R _residual for a 4x4 block is described below. Similar calculations may be performed for blocks of different sizes. Encoding module 30 calculates a matrix of transform coefficients for the residual data. An example matrix of transform coefficients is shown below.

The number of rows of the matrix A of transform coefficients is equal to the number of rows of pixels in the block and the number of columns of the matrix of transform coefficients is equal to the number of columns of pixels in the block. As such, in the above example, the dimension of the matrix of changed coefficients is 4 × 4 corresponding to a 4 × 4 block. Each of entries A (i, j) of the matrix of transform coefficients is a transform of respective residual coefficients.

During quantization, the transform coefficients of matrix A with smaller values tend to be zero after quantization. As such, encoding module 30 compares matrix A of residual transform coefficients with a matrix of thresholds to predict which transform coefficient of matrix A will remain non-zero after quantization. An example matrix of thresholds is shown below.

The matrix C may be calculated as a function of the QP value. The dimension of matrix C is the same as the dimension of matrix A. In the case of the H.264 standard, for example, the entry of matrix C may be calculated based on equation (5).

QBITS {QP} is a parameter that determines scaling as a function of QP, and Level-Offset (i, j) {QP} is a dead zone parameter for entries in row i and column j of the matrix, and also a function of QP , Level_Scale (i, j) {QP} is the multiplication factor for the entries in row i and column j of the matrix and is a function of QP, i is the row of the matrix, j is the column of the matrix and QP is the encoding Corresponds to the quantization parameter of module 30. In the exemplary equation (5), the variable may be defined in the H.264 coding standard as a function of operation QP. Other equations may be used to determine which of the variables will remain in quantization, and other equations may be defined in other coding standards based on the quantization method adopted by a particular standard. In some aspects, encoding module 30 may be configured to operate within a range of QP values. In this case, encoding module 30 may precalculate a plurality of comparison matrices corresponding to each of the QP values within the range of QP values. Encoding module 30 selects a comparison matrix corresponding to the QP of encoding module 30 and compares it with the transform coefficient matrix.

The result of the comparison between the matrix A of the transform coefficients and the matrix C of the threshold is a matrix of ones and zeros. In the above example, this comparison is a matrix of 1s and 0s shown below:

1 indicates the position of transform coefficients that are likely to remain in quantization, i.e., remain non-zero, and 0 indicates the position of transform coefficients that are unlikely to remain in quantization, i.e., become zero. Indicates. As mentioned above, the transform coefficients are identified to remain non-zero when the absolute value of the transform coefficients of matrix A is greater than or equal to the corresponding threshold of matrix C.

Using the resulting matrix of 1s and 0s, bit estimation module 42 determines the number of transform coefficients that will remain in quantization. In other words, bit estimation module 42 determines the number of transform coefficients identified as being left non-zero after quantization. Bit estimation module 42 may determine the number of transform coefficients identified as remaining non-zero after quantization according to equation (6).

NZ _est is an estimated number of non-zero transform coefficients, and M (i, j) is the value of matrix M in row i and column j. In the above example, NZ _est is equal to eight.

Bit estimation module 42 also calculates the sum of at least some of the absolute values of the transform coefficients estimated to remain in quantization. In a particular aspect, the bit estimation module 42 may calculate the sum of at least some of the absolute values of the transform coefficients according to equation (7):

SATD is the sum of transform coefficients identified as remaining non-zero after quantization, M (i, j) is the value of matrix M in row i and column j, and A (i, j) is row i and column The value of matrix A in j, abs (x) is an absolute value function that computes the absolute value of x. In the example described above, the SATD is equal to 2361. Like SSTDs, other different matrices may be used for the transform coefficients.

Using these values, bit estimation module 42 approximates the number of bits associated with the coding of residual coefficients using equation (3) above. Control module 32 may use the estimate of R _residual to calculate an estimate of the total coding cost of the mode. Encoding module 30 may estimate the total coding cost for one or more other possible modes in the same manner, and then select the mode with the minimum coding cost. Encoding module 30 then applies the selected coding mode to code the block or blocks of the frame.

The foregoing techniques may be implemented separately, or two or more of the techniques, or both, may be implemented together in the encoding device 12. Components in encoding module 30 are examples of components applicable for implementing the techniques described herein. However, encoding module 30 may include many other components, as well as fewer components that combine the functionality of one or more modules described above, if desired. The components of encoding module 30 may be implemented as one or more processors, digital signal processors, application specific integrated semiconductors (ASICs), field programmable gate arrays (FPGAs), separate logic, software, hardware, firmware, or any combination thereof. have. The description of the different features as a module is intended to highlight different functional aspects of the encoding module 30 and does not necessarily mean that these modules must be realized by separating hardware or software components. Rather, functionality associated with one or more modules may be integrated within common or separate hardware or software components.

3 is a block diagram illustrating another example encoding module 50. Encoding module 50 of FIG. 3 uses the method of FIG. 2 except that bit estimation module 52 of encoding module 50 estimates the number of bits associated with coding of the residual data after quantization of the transform coefficients for the residual data. Substantially consistent with encoding module 30. In particular, after quantization of the transform coefficients, bit estimation module 52 uses equation (8) to estimate the number of bits associated with the coding of the residual coefficients:

SATQD is the sum of absolute values of the non-zero quantized transform coefficients, NZ _TQ is the number of non-zero quantized transform coefficients, and a ₁ , a _2, and a ₃ are coefficients. The coefficients a ₁ , a _2, and a ₃ may be calculated using, for example, a least squares estimate. Even though encoding module 50 quantizes the transform coefficients before estimating the number of bits associated with coding of the residual data, encoding module 50 still estimates the coding cost for the modes without actually estimating the data of the blocks. . Thus, the computationally intensive computation amount is still reduced.

4 is a flow diagram illustrating exemplary operations of an encoding module, such as encoding module 30 of FIG. 2 and / or encoding module 50 of FIG. 3, selecting an encoding module based at least on an estimated coding cost. . However, for illustration purposes, FIG. 4 will discuss the encoding module 30. Encoding module 30 selects a mode for estimating coding cost 60 (60). Encoding module 30 generates a distortion metric for the current block (62). For example, encoding module 30 may calculate a distortion metric based on the comparison between the block and the at least one reference block. For a block selected to be intra coded, the reference block may be an adjacent block within the same frame. On the other hand, for a block selected to be inter coded, the reference block may be a block from an adjacent block. The distortion metric may be, for example, a SAD, SSD, SATD, SSTD or other similar distortion metric.

In the example of FIG. 4, encoding module 30 uses a non-context adaptive coding technique to determine the number of bits associated with coding of a portion of the coded data (64). As described above, this data may include one or more motion vectors of the block, an identifier indicating the coding mode of the block, one or more reference frame indices, QP information, slice information of the block, and the like. Encoding module 30 may use a code table to identify the number of bits associated with coding of data using, for example, FLC, universal VLC, or context adaptive coding techniques.

Encoding module 30 estimates and / or calculates the number of bits associated with the coding of the portion of the coded data using the context adaptive coding technique (66). In the context of the H.264 standard, for example, encoding module 30 may estimate the number of bits associated with coding of residual data using context adaptive coding. Encoding module 30 may estimate the number of bits associated with coding of residual data without actually performing coding of residual data. In a particular aspect, encoding module 30 may estimate the number of bits associated with coding of the residual data without quantizing the residual data. For example, encoding module 30 may calculate transform coefficients for the residual data and identify transform coefficients that are likely to remain non-zero after quantization. Using these identified transform coefficients, encoding module 30 estimates the number of bits associated with the coding of the residual data. In another aspect, encoding module 30 may quantize the transform coefficients and estimate the number of bits associated with coding of residual data based at least on the quantized transform coefficients. In either case, encoding module 30 saves time and processing resources by estimating the required number of bits. If there are sufficient computing resources, encoding module 30 may calculate the actual number of bits required instead of estimation.

Encoding module 30 estimates and / or calculates the total coding cost for coding the block in the selected mode (68). Encoding module 30 is based on the distortion metric, the bits associated with the coding of the portion of the data coded using non-context adaptive coding, and the bits associated with the coding of the portion of the data coded using context adaptive coding. To estimate the total coding cost for coding the block. For example, encoding module 30 may estimate the total coding cost for coding the block in the selected mode using equation (2) or equation (3) above.

Encoding module 30 determines whether there are any other coding modes for estimating coding cost (70). As described above, encoding module 30 estimates the coding cost for at least some of the possible modes. In a particular aspect, encoding module 30 may estimate the coding cost of the block in each of the possible coding modes. In the context of the H.264 standard, for example, encoding module 30 is configured for 22 different coding modes (inter coding mode and intra coding mode) for a block selected for inter coding and for a block selected for intra coding. Coding costs may be estimated for 13 different coding modes. In other aspects, encoding module 30 may use another mode selection technique to initially reduce the set of possible modes, and thereafter, to estimate the coding cost for the reduced set of coding modes. Technology can also be used.

When there are further coding modes for estimating the coding cost, encoding module 30 selects the next coding mode and estimates the coding cost of the data in the selected coding mode. When a coding mode for estimating coding cost no longer exists, encoding module 30 selects one of the modes to use for coding of the block based at least on the estimated coding cost (72). In one example, coding module 30 may select a coding mode with the least estimated coding cost. In selecting a mode, coding module 30 may apply the selected mode to code a particular block (74). This process may continue for additional blocks within a given frame. By way of example, the process may continue until all the blocks in the frame are coded using the coding mode selected in connection with the techniques described herein. This process may also continue until blocks of a plurality of frames are coded using a high efficiency mode.

5 is a flowchart illustrating an exemplary operation of an encoding module, such as encoding module 30 of FIG. 2, to estimate the number of bits associated with coding of residual coefficients of the block. After selecting one of the coding modes to estimate the coding cost, encoding module 30 generates residual data of the block for the selected mode (80). For a block selected to be intra coded, for example, spatial prediction module 34 generates residual data for that block based on the comparison of the block with the predicted version of the block. Alternatively, for a block selected to be inter coded, motion estimation module 36 and motion compensation module 38 calculate residual data for that block based on the comparison between that block and the corresponding block in the reference frame. In some aspects, the residual data may have already been calculated to generate the distortion metric of the block. In this case, encoding module 30 may retrieve the residual data from the memory.

Transform module 40 transforms the residual coefficients of the block in accordance with the transform function to generate transform coefficients for the residual data (82). Transform module 40 may apply a 4x4 or 8x8 integer transform or a DCT transform to the residual data, for example, to generate transform coefficients for the residual data. Bit estimation module 42 compares one of the transform coefficients with a corresponding threshold to determine whether the transform coefficient is greater than or equal to the threshold (84). The threshold corresponding to the transform coefficients may be calculated as a function of QP of encoding module 30. If the transform coefficient is greater than or equal to the corresponding threshold, bit estimation module 42 identifies the transform coefficient as a coefficient that will remain non-zero after quantization (86). If the transform coefficient is smaller than the corresponding threshold, bit estimation module 42 identifies the transform coefficient as a coefficient that will be zero after quantization (88).

Bit estimation module 42 determines whether there are any additional transform coefficients for the residual data of the block (90). If there are additional transform coefficients of the block, bit estimation module 42 selects the other one of the coefficients and compares it with the corresponding threshold. If there are no additional transform coefficients to analyze, bit estimation module 42 determines the number of coefficients identified as remaining non-zero after quantization (92). Bit estimation module 42 also sums at least some of the absolute values of the transform coefficients identified as remaining non-zero after quantization (94). Bit estimation module 42 estimates the number of bits associated with the coding of the residual data using the sum of the determined number of non-zero coefficients and a portion of the non-zero coefficients (96). Bit estimation module 42 may, for example, estimate the number of bits associated with coding of residual data using Equation (4) above. In this way, encoding module 30 estimates the number of bits associated with the coding of the residual data of the block in the selected mode, without quantizing or encoding the residual data.

FIG. 6 is a flowchart illustrating an exemplary operation of an encoding module, such as encoding module 50 of FIG. 3, to estimate the number of bits associated with coding of residual coefficients of a block. After selecting one of the coding modes for estimating the coding cost, encoding module 50 generates residual coefficients of the block (100). For a block selected to be intra coded, for example, spatial prediction module 34 calculates residual data for that block based on the comparison of the block with the predicted version of the block. Alternatively, for a block selected to be inter coded, motion estimation module 36 and motion compensation module 38 calculate residual data for that block based on the comparison between that block and the corresponding block in the reference frame. In some aspects, the residual coefficients may have already been calculated to produce a distortion metric of the block.

Transform module 40 transforms the residual coefficients of the block according to the transform function to generate transform coefficients for the residual data (102). Transform module 40 may apply a 4x4 or 8x8 integer transform or a DCT transform to the residual data, for example, to produce transformed residual coefficients. Quantization module 46 quantizes the transform coefficients according to the QP of encoding module 50 (104).

Bit estimation module 52 determines the number of non-zero quantized transform coefficients (106). Bit estimation module 42 also sums the absolute values of the non-zero levels or quantized transform coefficients (108). Bit estimation module 52 estimates the number of bits associated with coding of residual data using the sum of the calculated number of non-zero quantized transform coefficients and the non-zero quantized transform coefficients (110). Bit estimation module 52 may, for example, estimate the number of bits associated with the coding of residual coefficients using equation (4) above. In this way, the encoding module estimates the number of bits associated with the coding of the residual data of the block in the selected mode without encoding the residual data.

 Based on the teachings described herein, it is apparent that an aspect disclosed herein may be implemented independently of any other aspect, and two or more aspects may be combined in various ways. The techniques described herein may be implemented in hardware, software, firmware, or any combination thereof. If implemented in hardware, the present teachings may be implemented using digital hardware, analog hardware, or a combination thereof. If implemented in software, the techniques may be implemented at least in part by a computer program product including a computer readable medium having stored instructions and code. The instructions or code associated with the computer readable medium of the computer program product may be a computer, for example, one or more digital signal processors (DSPs), general purpose microprocessors, ASICs, FPGAs, or other equivalent integrated or separate logic circuits. It may be implemented by one or more processors, such as.

By way of example, computer-readable media may comprise RAM, such as synchronous dynamic RAM (SDRAM), read only memory (ROM), nonvolatile random access memory (NVRAM), ROM, electrically erasable and programmable read only memory (EEPROM), EEPROM, FLASH memory, CD-ROM or other optical disk storage, magnetic disk storage or other magnetic storage device, or can be used to transfer or store the desired program code in the form of instructions and data structures and can be accessed by a computer It may include, but is not limited to any other tangible media.

Numerous aspects and embodiments have been described. However, various modifications are possible to these embodiments, and likewise, the principles presented herein may be applied to other aspects. These and other aspects are within the scope of the following claims.

Claims (40)

As a method of processing digital video data, Identifying, for each of the plurality of coding modes, one or more transform coefficients for the residual data of the block of pixels that will remain non-zero when quantized, wherein the one or more transform coefficients are non-zero and unquantized transform coefficients Identifying the transform coefficients; Estimating, for each of the plurality of coding modes, the number of bits associated with the coding of the residual data in each of the plurality of coding modes based on at least the identified transform coefficients for each of the plurality of coding modes; Based on the estimated number of bits associated with the coding of the residual data in each of the plurality of coding modes, coding cost for coding the block of pixels in each of the plurality of coding modes, the plurality of codings. Estimating for each mode; And Selecting and applying a coding mode having the smallest coding cost of the estimated coding costs among the plurality of coding modes to code the block of pixels, Estimating the number of bits associated with coding of the residual data is performed without quantizing the identified non-zero quantized transform coefficients. The method of claim 1, Identifying the transform coefficients includes comparing each of the transform coefficients to a corresponding one of a plurality of thresholds to identify transform coefficients that will remain non-zero when quantized, wherein the plurality of thresholds Each of which is calculated as a function of a quantization parameter (QP). The method of claim 2, Comparing each of the transform coefficients with a corresponding one of a plurality of thresholds, identifying the transform coefficients that will remain non-zero when quantized is the transform coefficients that will remain non-zero when quantized, corresponding to Identifying transform coefficients that are greater than or equal to a threshold. The method of claim 2, Precomputing a plurality of sets of thresholds, each of the sets of thresholds corresponding to a different value of the QP; And Selecting one of the plurality of sets of thresholds based on the value of the QP used to encode the block of pixels. The method of claim 1, Estimating the number of bits associated with the coding of the residual data, Determining the number of transform coefficients identified as being left non-zero when quantized; Summing the absolute values of at least one of the transform coefficients identified as being left non-zero when quantized; And Estimating the number of bits associated with the coding of the residual data based at least on the sum of the determined number of non-zero transform coefficients and the absolute values of the at least one non-zero transform coefficient. Treatment method. delete delete The method of claim 1, Estimating the coding cost, Calculating a distortion metric for the block of pixels; Calculating a number of bits associated with coding of non-residual data of the block of pixels; And Estimating a coding cost for coding the block of pixels based at least on the distortion metric, the number of bits associated with coding of the non-residual data, and the number of bits associated with coding of the residual data, Method of processing digital video data. The method of claim 1, Quantizing the transform coefficients for the residual data after the step of selecting the coding mode; Encoding the quantized transform coefficients for the residual data; And Transmitting the encoded coefficients for the residual data. The method of claim 1, Generating a matrix of transform coefficients, wherein the number of rows of the matrix of transform coefficients is equal to the number of rows of pixels in the block, and the number of columns of the matrix of transform coefficients is of the column of pixels in the block. Generating a matrix of transform coefficients, such as a number; Comparing the matrix of transform coefficients to a matrix of thresholds, the matrix of thresholds having the same dimensions as the matrix of transform coefficients, the comparison also being a matrix of ones and zeros, wherein zero is zero after quantization Comparing the matrix of transform coefficients to a matrix of thresholds, wherein the matrix represents a position in the matrix of transform coefficients to be represented and wherein 1 represents a position in the matrix of transform coefficients to remain non-zero after quantization; Summing the number of 1s in the matrix of 1s and 0s to calculate the number of transform coefficients identified as being non-zero when quantized; Summing absolute values of at least one of the transform coefficients corresponding to the position of 1 in the matrix of 1 and 0 in the matrix of transform coefficients; And Estimating the number of bits associated with coding of the residual data based at least on the number of non-zero transform coefficients and the sum of the at least one non-zero transform coefficient. . An apparatus for processing digital video data, Means for identifying, for each of the plurality of coding modes, one or more transform coefficients for the residual data of the block of pixels that will remain non-zero when quantized; Means for estimating, for each of the plurality of coding modes, the number of bits associated with the coding of the residual data in each of the plurality of coding modes based on at least the identified transform coefficients for each of the plurality of coding modes; And Based on the estimated number of bits associated with the coding of the residual data in each of the plurality of coding modes, coding cost for coding the block of pixels in each of the plurality of coding modes, the plurality of codings. Means for estimating for each mode; And Means for selecting and applying a coding mode having the smallest coding cost of an estimated coding cost among the plurality of coding modes to code the block of pixels, And estimating the number of bits associated with the coding of the residual data is performed prior to quantization of any of the at least the identified transform coefficients. The method of claim 11, wherein A transform module for generating transform coefficients for the residual data of the block of pixels, The means for identifying the one or more transform coefficients for each of the plurality of coding modes and the means for estimating the number of bits for each of the plurality of coding modes comprise the one or more transform coefficients that will remain non-zero when quantized. Identify for each of the plurality of coding modes and determine the number of bits associated with the coding of the residual data in each of the plurality of coding modes based on at least the identified transform coefficients for each of the plurality of coding modes, A bit estimation module for estimating for each of the plurality of coding modes, The means for estimating the coding cost for each of the plurality of coding modes is based on at least an estimated number of bits associated with the coding of the residual data in each of the plurality of coding modes in each of the plurality of coding modes. And a control module for estimating a coding cost for coding the block of pixels of for each of the plurality of coding modes. 13. The method of claim 12, The bit estimation module compares each of the transform coefficients with a corresponding one of a plurality of thresholds to identify transform coefficients that will remain non-zero when quantized, each of the plurality of thresholds being a quantization parameter (QP). A device for processing digital video data, calculated as a function. The method of claim 13, And the bit estimation module identifies transform coefficients that are less than a corresponding threshold as transform coefficients that will remain non-zero when quantized. The method of claim 13, The bit estimation module precalculates a plurality of sets of thresholds, each of the sets of thresholds corresponding to a different value of the QP, and based on the value of the QP used to encode the block of pixels. And select one of the plurality of sets. 13. The method of claim 12, The bit estimation module determines the number of the transform coefficients identified as remaining non-zero when quantized, sums the absolute values of at least one of the transform coefficients identified as remaining non-zero when quantized, And estimating the number of bits associated with the coding of the residual data based at least on the sum of the determined number of non-zero transform coefficients and the absolute values of the at least one non-zero transform coefficient. Device. delete delete 13. The method of claim 12, The control module calculates a distortion metric for the block of pixels, calculates the number of bits associated with the coding of non-residual data of the block of pixels, and at least is associated with the distortion metric, coding of the non-residual data. And estimate a coding cost for coding the block of pixels based on the number of bits and the number of bits associated with the coding of the residual data. 13. The method of claim 12, A quantization module for quantizing the transform coefficients for the residual data after selecting the coding mode; An entropy encoding module for encoding the quantized transform coefficients for the residual data; And And a transmitter for transmitting the encoded coefficients for the residual data. 13. The method of claim 12, The transform module generates the matrix of transform coefficients, wherein the number of rows of the matrix of transform coefficients is equal to the number of rows of pixels in the block, and the number of columns of the matrix of transform coefficients is the column of pixels in the block. Equal to, and The bit estimation module compares the matrix of transform coefficients with a matrix of thresholds, the matrix of thresholds having the same dimensions as the matrix of transform coefficients, and wherein the comparison is a matrix of 1s and 0s, where 0 is Indicates a position in the matrix of transform coefficients that will be zero after quantization and 1 denotes a position in the matrix of transform coefficients that will remain non-zero after quantization, The bit estimation module also calculates the number of transform coefficients identified as remaining non-zero when quantized by summing the number of ones in the matrix of ones and zeros, and calculating the number of ones in the matrix of transform coefficients. Sum the absolute values of at least one of the transform coefficients corresponding to the position of 1 in the matrix of zero, and based at least on the sum of the number of non-zero transform coefficients and the at least one non-zero transform coefficient And estimating the number of bits associated with the coding of the residual data. The method of claim 11, wherein The identifying means compares each of the transform coefficients with a corresponding one of a plurality of thresholds to identify transform coefficients that will remain non-zero when quantized, each of the plurality of thresholds being a quantization parameter (QP). A device for processing digital video data, calculated as a function. 23. The method of claim 22, And said identifying means identifies transform coefficients that are less than a corresponding threshold as transform coefficients that will remain non-zero when quantized. 23. The method of claim 22, Means for precomputing a plurality of sets of thresholds, each of the sets of thresholds corresponding to a different value of the QP; And And means for selecting one of the plurality of sets of thresholds based on a value of a QP used to encode the block of pixels. 22. The method of claim 21, The bit estimating means, Determine the number of transform coefficients identified as remaining non-zero when quantized, sum the absolute values of at least one of the transform coefficients identified as remaining non-zero when quantized, and, at least, the non- And estimate the number of bits associated with the coding of the residual data based on the sum of the determined number of zero transform coefficients and the absolute values of the at least one non-zero transform coefficient. delete delete 22. The method of claim 21, The coding cost estimation means calculates a distortion metric for the block of pixels, calculates a number of bits associated with coding of non-residual data of the block of pixels, and at least, the distortion metric, the non-residual data. And estimate a coding cost for coding the block of pixels based on the number of bits associated with a coding of and the number of bits associated with coding of the residual data. 22. The method of claim 21, Means for quantizing the transform coefficients for the residual data after selecting the coding mode; Means for encoding the quantized transform coefficients for the residual data; And And means for transmitting the encoded coefficients for the residual data. 22. The method of claim 21, Means for generating a matrix of transform coefficients, wherein the number of rows of the matrix of transform coefficients is equal to the number of rows of pixels in the block, and the number of columns of the matrix of transform coefficients is the number of columns of pixels in the block Means for generating a matrix of transform coefficients, such as The identifying means compares the matrix of transform coefficients with a matrix of thresholds, the matrix of thresholds having the same dimensions as the matrix of transform coefficients, the comparison is a matrix of 1s and 0s, and 0 is quantized A position in the matrix of transform coefficients to be zero afterwards and 1 denotes a position in the matrix of transform coefficients to be left non-zero after quantization; And The bit estimating means calculates the number of transform coefficients identified as remaining non-zero when quantized by summing the number of 1s in the matrix of 1s and 0s, and calculating the number of transform coefficients 1 and 0 in the matrix of transform coefficients. Summing the absolute values of at least one of the transform coefficients corresponding to the position of 1 in the matrix, and based at least on the sum of the number of non-zero transform coefficients and the at least one non-zero transform coefficient And estimating the number of bits associated with the coding of residual data. A computer readable medium having instructions for processing digital video data, comprising: The commands are Code for identifying, for each of the plurality of coding modes, one or more transform coefficients for the residual data of the block of pixels that will remain non-zero when quantized; Code for estimating, for each of the plurality of coding modes, the number of bits associated with the coding of the residual data in each of the plurality of coding modes based on at least the identified transform coefficients for each of the plurality of coding modes. ; Based on the estimated number of bits associated with the coding of the residual data in each of the plurality of coding modes, coding cost for coding the block of pixels in each of the plurality of coding modes, the plurality of codings. Code for estimating for each mode; And Code for selecting and applying a coding mode having the smallest coding cost among estimated coding costs among the plurality of coding modes, for coding the block of pixels, Estimating the number of bits associated with coding of the residual data is performed without quantizing the at least the identified transform coefficients. 32. The method of claim 31, The code identifying the transform coefficients includes a code identifying each of the transform coefficients with a corresponding one of a plurality of thresholds to identify transform coefficients that will remain non-zero when quantized, wherein the plurality of thresholds Each calculated as a function of quantization parameter (QP). 33. The method of claim 32, Comparing each of the transform coefficients with a corresponding one of a plurality of thresholds, the code identifying the transform coefficients that will remain non-zero when quantized is the transform coefficients that will remain non-zero when quantized, corresponding to And code for identifying transform coefficients less than a threshold. 33. The method of claim 32, Code for precomputing a plurality of sets of thresholds, each of the sets of thresholds corresponding to a different value of the QP; And And code for selecting one of the plurality of sets of thresholds based on a value of a QP used to encode the block of pixels. 32. The method of claim 31, A code for estimating the number of bits associated with coding of the residual data is Code for determining the number of transform coefficients identified as being left non-zero when quantized; Code for summing the absolute values of at least one of the transform coefficients identified as being left non-zero when quantized; And And code for estimating the number of bits associated with coding of the residual data based at least on the sum of the determined number of non-zero transform coefficients and the absolute values of the at least one non-zero transform coefficient. media. delete delete 32. The method of claim 31, The code for estimating the coding cost, Code for calculating a distortion metric for the block of pixels; Code for calculating the number of bits associated with coding of non-residual data of the block of pixels; And Code for estimating a coding cost for coding the block of pixels based at least on the distortion metric, the number of bits associated with coding of the non-residual data, and the number of bits associated with coding of the residual data, Computer readable media. 32. The method of claim 31, Code for quantizing transform coefficients for the residual data after selecting the coding mode; Code for encoding the quantized transform coefficients for the residual data; And And code for transmitting the encoded coefficients for the residual data. 32. The method of claim 31, Code for generating the matrix of transform coefficients, wherein the number of rows of the matrix of transform coefficients is equal to the number of rows of pixels in the block, and the number of columns of the matrix of transform coefficients is the number of columns of pixels in the block Code for generating a matrix of transform coefficients, such as; Code for comparing the matrix of transform coefficients to a matrix of thresholds, the matrix of thresholds having the same dimensions as the matrix of transform coefficients, wherein the comparison is a matrix of ones and zeros, and zero is zero after quantization Code for comparing the matrix of transform coefficients to a matrix of thresholds, wherein the position represents a position in the matrix of transform coefficients to be represented and wherein 1 is a position in the matrix of transform coefficients to remain non-zero after quantization; Summing a number of ones in the matrix of ones and zeros to calculate the number of transform coefficients identified as being non-zero when quantized; Code for summing absolute values of at least one of the transform coefficients corresponding to the position of 1 in the matrix of 1 and 0 in the matrix of transform coefficients; And And code for estimating the number of bits associated with coding of the residual data based at least on the number of non-zero transform coefficients and the sum of the at least one non-zero transform coefficients.

Images