Hubbry Logo
MacroblockMacroblockMain
Open search
Macroblock
Community hub
Macroblock
logo
7 pages, 0 posts
0 subscribers
Be the first to start a discussion here.
Be the first to start a discussion here.
Contribute something
Macroblock
Macroblock
Macroblock
View on Wikipedia
from Wikipedia

The macroblock is a processing unit in image and video compression formats based on linear block transforms, typically the discrete cosine transform (DCT). A macroblock typically consists of 16×16 samples, and is further subdivided into transform blocks, and may be further subdivided into prediction blocks. Formats which are based on macroblocks include JPEG, where they are called MCU blocks, H.261, MPEG-1 Part 2, H.262/MPEG-2 Part 2, H.263, MPEG-4 Part 2, and H.264/MPEG-4 AVC.[1][2][3][4] In H.265/HEVC, the macroblock as a basic processing unit has been replaced by the coding tree unit.[5]

Technical details

[edit]

Transform blocks

[edit]

A macroblock is divided into transform blocks, which serve as input to the linear block transform, e.g. the DCT. In H.261, the first video codec to use macroblocks, transform blocks have a fixed size of 8×8 samples.[1] In the YCbCr color space with 4:2:0 chroma subsampling, a 16×16 macroblock consists of 16×16 luma (Y) samples and 8×8 chroma (Cb and Cr) samples. These samples are split into four Y blocks, one Cb block and one Cr block. This design is also used in JPEG and most other macroblock-based video codecs with a fixed transform block size, such as MPEG-1 Part 2 and H.262/MPEG-2 Part 2. In other chroma subsampling formats, e.g. 4:0:0, 4:2:2, or 4:4:4, the number of chroma samples in a macroblock will be smaller or larger, and the grouping of chroma samples into blocks will differ accordingly.

In more modern macroblock-based video coding standards such as H.263 and H.264/AVC, transform blocks can be of sizes other than 8×8 samples. For instance, in H.264/AVC main profile, the transform block size is 4×4.[4] In H.264/AVC High profile, the transform block size can be either 4×4 or 8×8, adapted on a per-macroblock basis.[4]

Prediction blocks

[edit]

Distinct from the division into transform blocks, a macroblock can be split into prediction blocks. In early standards such as H.261, MPEG-1 Part 2, and H.262/MPEG-2 Part 2, motion compensation is performed with one motion vector per macroblock.[1][2] In more modern standards such as H.264/AVC, a macroblock can be split into multiple variable-sized prediction blocks, called partitions.[4] In an inter-predicted macroblock in H.264/AVC, a separate motion vector is specified for each partition.[4] Correspondingly, in an intra-predicted macroblock, where samples are predicted by extrapolating from the edges of neighboring blocks, the predicted direction is specified on a per-partition basis.[4] In H.264/AVC, prediction partition size ranges from 4×4 to 16×16 samples for both inter-prediction (motion compensation) and intra-prediction.[4]

Bitstream representation

[edit]

A possible bitstream representation of a macroblock in a video codec which uses motion compensation and transform coding is given below.[6] It is similar to the format used in H.261.[1] 

+------+------+-------+--------+-----+----+----+--------+
| ADDR | TYPE | QUANT | VECTOR | CBP | b0 | b1 | ... b5 |
+------+------+-------+--------+-----+----+----+--------+
  • ADDR — address of block in image
  • TYPE — identifies type of macroblock (intra frame, inter frame, bi-directional inter frame)
  • QUANT — quantization value to vary quantization
  • VECTOR — motion vector
  • CBP — Coded Block Pattern, this is bit mask indicating for which blocks coefficients are present.
  • bN — the blocks (4 Y, 1 Cr, 1 Cb)

Macroblocking

[edit]
Main article: Compression artifact

The term macroblocking is commonly used to refer to block coding artifacts.

See also

[edit]

References

[edit]
Revisions and contributorsEdit on WikipediaRead on Wikipedia

Macroblock

View on Grokipedia
from Grokipedia
A macroblock is the fundamental processing unit in block-based video compression standards, consisting of a 16×16 block of luminance (luma) samples and two corresponding 8×8 blocks of chrominance (chroma) samples in the common 4:2:0 color subsampling format. This structure enables efficient handling of spatial redundancy through techniques such as motion estimation, compensation, intra- and inter-prediction, and application of the discrete cosine transform (DCT) for quantization and encoding. Macroblocks form the basis of hybrid coding schemes that divide video frames into a grid for processing, allowing for reduced bitrate while maintaining perceptual quality in applications like streaming, broadcasting, and storage. The macroblock concept originated in the ITU-T H.261 standard, released in 1990 as the first practical digital video codec for low-bitrate videoconferencing over ISDN lines, where it was defined as a 16×16 luma block with 8×8 chroma components to support block-matching motion compensation. It was subsequently adopted and refined in successive standards, including MPEG-1 (1992) for CD-ROM video, MPEG-2 (1995) for DVD and digital TV, H.263 (1996) for improved low-bitrate internet video, and MPEG-4 Part 2 (1999) for object-based coding. The most widespread evolution occurred in H.264/AVC (Advanced Video Coding, 2003), a joint ITU-T and ISO/IEC standard that enhanced macroblock flexibility by allowing partitions into smaller sub-blocks (down to 4×4) for more precise motion vectors and transform sizes, achieving up to 50% bitrate savings over prior codecs. In modern video coding, the fixed 16×16 macroblock has largely been supplanted by more adaptive structures, such as the coding tree units (CTUs) in HEVC/H.265 (2013), which support larger blocks up to 64×64 pixels for better efficiency in high-resolution content like 4K and 8K video. Nonetheless, macroblocks remain relevant in legacy systems, ongoing H.264 deployments, and as a foundational element influencing block-based partitioning in emerging standards like VVC/H.266 (2020), where they inform hierarchical coding decisions for ultra-high-definition and immersive media.

Fundamentals

Definition and Purpose

A macroblock serves as the fundamental processing unit in block-based video codecs, such as those defined in the ITU-T H.261 and H.264 standards. It typically comprises a 16×16 array of luma samples, along with associated chroma samples—such as two 8×8 arrays for the Cb and Cr components in 4:2:0 color sampling formats. This structure allows the macroblock to represent a compact spatial region within a video frame, facilitating localized analysis and manipulation of pixel data during compression. The primary purpose of the macroblock is to enable efficient spatial and temporal compression by grouping pixels into discrete units suitable for , intra- and inter-prediction, and . In , for instance, the macroblock is matched against reference blocks from previous or future frames to compute motion vectors, exploiting temporal redundancy across video sequences. Similarly, spatial prediction within the macroblock leverages adjacent pixel correlations to minimize residual data, which is then transformed (e.g., via ) and quantized to further reduce bitrate while preserving essential visual information. This block-based approach originated in early standards like for videoconferencing and has been refined in subsequent codecs to achieve higher compression ratios. Key benefits of using macroblocks include simplified in encoding and decoding pipelines, as operations are confined to fixed-size blocks rather than processing the entire frame holistically, which optimizes hardware and software implementations. This partitioning enhances overall compression efficiency by allowing adaptive techniques, such as variable block partitioning for better accuracy, leading to improved video quality at lower bitrates without excessive computational overhead. In standard-definition video (e.g., 720×480 resolution), a single macroblock might cover a small detail like part of a face or a uniform background patch, demonstrating its role in balancing detail preservation with data reduction.

Historical Development

The macroblock concept emerged in the late amid the development of early block-based video codecs, addressing the need for efficient compression in bandwidth-limited telecommunication applications. It was first formalized in the standard, ratified in 1990 for video telephony over ISDN lines at bitrates ranging from 64 to 2048 kbit/s. specified the macroblock as a 16×16 luma block accompanied by corresponding 8×8 chroma blocks, serving as the fundamental unit for differenced inter-frame coding through and . This structure set the template for subsequent standards, enabling practical transmission in resource-constrained environments. Subsequent adoption expanded the macroblock's role across storage and broadcast media. The ISO/IEC standard, released in 1993, incorporated H.261's 16×16 macroblock framework for CD-ROM-based video at approximately 1 Mbps, introducing bi-directional prediction to enhance temporal redundancy reduction for resolutions like and . This was followed by (ITU-T H.262), standardized in 1995 through joint and MPEG collaboration, which retained the fixed 16×16 macroblock while adding interlaced-scan support for and DVD applications at 2–20 Mbps. These milestones reflected the era's hardware computational limitations, prioritizing algorithms that balanced efficiency with feasible real-time processing on 1990s-era processors. The early 2000s brought evolutionary refinements driven by rising demands for streaming and higher resolutions, amid persistent bandwidth constraints. The H.264/AVC standard, finalized in 2003 by the Video Coding Experts Group (VCEG) and ISO/IEC (MPEG) Joint Video Team, preserved the 16×16 macroblock as the processing unit but introduced variable-size partitions down to 4×4 for more adaptive , achieving roughly double the compression of prior standards. Building on this, the High Efficiency Video Coding (HEVC, ITU-T H.265) standard, approved in 2013, shifted from fixed macroblocks to larger Coding Tree Units (CTUs) up to 64×64 with recursive adaptive subdivisions, optimizing for HD and 4K video while further reducing bitrate needs by about 50% compared to H.264 at equivalent quality.

Technical Specifications

Macroblock Structure

A macroblock serves as the fundamental processing unit in video compression standards like H.264/AVC, comprising 256 luma samples arranged in a 16×16 grid, accompanied by chroma samples in the YCbCr color space. In the prevalent 4:2:0 chroma subsampling format, which is widely used for standard-definition and high-definition video, the macroblock includes two 8×8 blocks—one for the blue-difference (Cb) component and one for the red-difference (Cr) component—resulting in 64 chroma samples overall. This structure totals 384 samples per macroblock, calculated as: Total samples=256 (Y)+64 (Cb)+64 (Cr)=384\text{Total samples} = 256 \, (Y) + 64 \, (Cb) + 64 \, (Cr) = 384 The color space separates (Y) from (Cb and Cr), enabling efficient compression by exploiting human visual sensitivity to brightness over color details. H.264/AVC supports multiple subsampling ratios to accommodate varying applications: , where chroma resolution is quartered relative to luma (common in consumer video); 4:2:2, with horizontal by a factor of 2 (used in professional broadcast and editing workflows); and , preserving full chroma resolution (suited for high-fidelity graphics or ). For instance, in 4:2:2 format, each macroblock features 256 luma samples alongside 256 chroma samples (two 8×16 blocks of 128 samples each for Cb and Cr), resulting in 512 samples total. In format, the macroblock includes two 16×16 blocks for Cb and Cr, each with 256 samples (512 chroma samples total, 768 samples per macroblock). Macroblocks tile the video frame contiguously without overlap, ensuring complete coverage of the picture area. To facilitate this alignment, frame dimensions in luma samples are typically padded during preprocessing to multiples of 16 in both width and height, avoiding partial macroblocks at the edges. In progressive video, all samples within a macroblock are processed uniformly as a single spatial unit. For , however, the macroblock may be adaptively split into top-field and bottom-field components via macroblock-adaptive frame-field (MBAFF) coding, allowing independent processing of the interlaced lines to better handle motion artifacts.

Subdivisions and Blocks

In video coding standards such as H.264/AVC, a macroblock is subdivided into smaller blocks to enable more flexible and efficient processing for and transformation. These subdivisions allow the encoder to adapt to varying content characteristics, such as using larger blocks for uniform areas and smaller ones for detailed regions like edges. For inter , macroblocks can be partitioned into rectangular blocks including 16×16, 16×8, 8×16, and 8×8 sizes, with the 8×8 partitions further divisible into 8×4, 4×8, or 4×4 sub-partitions to refine . Intra , in contrast, operates on square blocks of 4×4 or 16×16 within the macroblock, facilitating directional spatial prediction. Transform blocks in H.264 are square and applied to the residual data after prediction, typically using 4×4 or 8×8 integer transforms akin to the (DCT). These block sizes balance computational efficiency with compression performance, with 4×4 blocks capturing high-frequency details in complex textures and 8×8 blocks handling smoother areas more effectively. The choice of subdivision is determined by the encoder's rate-distortion optimization, which selects partitions that minimize bitrate for a given quality level, often resulting in finer splits for high-motion or textured content. Building on H.264, the HEVC (H.265) standard evolves the macroblock concept into larger coding tree units (CTUs) of up to 64×64 pixels, which are recursively partitioned using a quad-tree structure down to minimum blocks of 4×4. This hierarchical approach allows for greater adaptability, where coding units (CUs) derived from CTU splits serve as the basis for prediction blocks ranging from 64×64 to 4×4, including non-square options like 16×8 for in irregular motion patterns. Transform blocks in HEVC extend to larger square sizes up to 32×32, also using DCT-like operations on residuals, enabling better energy compaction in high-resolution videos while maintaining finer granularity for detailed areas. The quad-tree partitioning promotes content-adaptive decisions, such as deeper splits around object edges to preserve sharpness without excessive bitrate overhead.
StandardPrediction Block ExamplesTransform Block Sizes
H.264/AVC16×16, 16×8, 8×16, 8×8, 4×44×4, 8×8
HEVC (H.265)64×64 to 4×4 (including rectangles like 16×8)4×4 to 32×32

Compression Processing

Prediction Mechanisms

Prediction mechanisms in macroblock-based video compression exploit redundancies to generate a predicted block, from which a residual is computed and encoded. In standards like H.264/AVC, macroblocks employ both intra and inter to minimize data transmission while preserving quality. The residual, defined as the difference between the current macroblock and its , is given by: residual=current macroblock−predicted macroblock\text{residual} = \text{current macroblock} - \text{predicted macroblock} This equation captures the prediction error, which is subsequently transformed and quantized. Intra prediction leverages spatial correlations within the same frame by estimating the current macroblock from neighboring encoded blocks. For luma components in H.264/AVC, two primary modes exist: intra 4×4 with 9 directional predictions (including vertical, horizontal, and diagonal modes) and DC prediction, applied to smaller blocks for textured regions; and intra 16×16 with 4 modes (vertical, horizontal, DC, and plane) suited for uniform areas. These modes use extrapolated or averaged samples from adjacent blocks, with unavailable boundary samples handled by replication or default values to ensure decoder synchronization. Chroma intra prediction follows a similar 16×16 structure with 4 modes. No prediction occurs across slice boundaries to maintain independence. Inter prediction utilizes temporal redundancies by matching the current macroblock to blocks in one or more reference frames via motion vectors, enabling efficient encoding of motion-dominated scenes. In P-slices, prediction draws from a single reference list with possible partitions down to 4×4 sub-blocks; B-slices support bi-prediction from two lists for enhanced accuracy. Motion vectors, typically at quarter-pixel precision for luma (using 6-tap interpolation for half-samples and bilinear for quarters), are differentially coded relative to predictors derived from neighboring vectors. Multiple reference frames (up to 16 in H.264/AVC) further improve by selecting the best match. Motion estimation identifies the optimal matching block in the reference frame by searching within a defined window, often using metrics like (SAD) to evaluate candidates. The process starts with integer-pixel searches (e.g., spiral or diamond patterns) centered on a predicted vector, followed by sub-pixel refinement for precision; costs incorporate both and vector bitrate via rate-distortion optimization, such as J=SAD+λ×mvbitsJ = \text{SAD} + \lambda \times \text{mvbits}, where λ\lambda scales with quantization parameter. A specialized skip mode optimizes static regions by directly copying the reference macroblock without transmitting residuals or explicit motion data, relying on a predicted zero or median motion vector. This mode, applicable in P- and B-slices, significantly reduces bitrate in unchanged scenes while signaling only the mode flag. Macroblocks may be subdivided into smaller partitions for refined inter prediction, as detailed in block subdivision techniques.

Transform Operations

In video compression standards such as and H.264/AVC, transform operations convert the residual data—obtained after —into the to facilitate efficient quantization and by concentrating energy in low-frequency coefficients. This process emphasizes lower frequencies, which typically represent the bulk of visual information, allowing higher-frequency components to be more aggressively discarded for compression. At the decoder, an inverse transform reconstructs the residual for addition to the predicted block. Early standards like (ITU-T H.262) employ the (DCT) on 8×8 blocks derived from macroblock residuals, enabling decorrelation and energy compaction suitable for quantization. In contrast, H.264/AVC introduces integer approximations of the DCT to avoid , reducing while maintaining performance close to the full DCT; the primary transform operates on 4×4 blocks for luma and chroma residuals, with an optional 8×8 transform available in high profiles for smoother areas. These integer transforms are designed as scaled versions of the DCT, ensuring exact invertibility in arithmetic. The H.264 transform is implemented separably using 1D operations along rows and columns, starting with a core 4-point 1D transform matrix for intra and inter residuals, followed by a 4-point on the DC coefficients of luma blocks in certain modes, and a 2×2 specifically for chroma DC coefficients to further compact low-frequency energy. This hierarchical approach enhances coding efficiency for chroma components, which often exhibit lower detail. Following transformation, the coefficients undergo quantization, where a quantization parameter (QP) scales the step size to control the trade-off between bitrate and quality; in H.264, QP ranges from 0 to 51, with each increment roughly doubling the quantization step and reducing bitrate by about 12% while increasing distortion. Coarser quantization (higher QP) discards more high-frequency details, potentially leading to blockier artifacts, whereas finer quantization preserves quality at higher bitrates. The foundational 1D DCT, approximated in these codecs, is defined for a block of length NN as: X=∑n=0N−1xcos⁡(π(2n+1)k2N),k=0,1,…,N−1X = \sum_{n=0}^{N-1} x \cos\left( \frac{\pi (2n+1) k}{2N} \right), \quad k = 0, 1, \dots, N-1 where xx are the input samples and XX are the coefficients, with modern approximations scaling factors to enable fixed-point computation.

Encoding Representation

Bitstream Syntax

In video compression standards such as and H.264, macroblocks are encoded within a hierarchical structure organized into frames or fields, which are further divided into slices for resilience and parallel processing. Each slice contains a sequence of macroblocks processed in raster-scan order, with syntax elements including headers that specify the macroblock type (intra-coded I, predicted P, or bi-predicted B), partition information for , and associated data such as residuals and motion parameters. This structure ensures that decoders can reconstruct the video by parsing the sequentially, starting from picture-level headers and progressing to macroblock-level details within each slice. Key syntax elements for macroblocks include the macroblock address increment (often abbreviated as MBAI or MBAP in older standards), which signals the relative position of the current macroblock to the previous one, typically using a variable-length code to skip over unchanged areas efficiently. Another critical element is the coded block pattern (CBP), a that indicates which of the macroblock's or smaller blocks contain non-zero quantized transform coefficients after , allowing decoders to skip irrelevant blocks and reduce bitstream overhead. For example, in H.264, the CBP is encoded as a 6-bit value for luma and chroma components, derived from the residual data presence in up to four luma and two chroma blocks. Entropy coding in modern standards like H.264 employs variable-length techniques to represent macroblock modes, motion vectors, and transform coefficients compactly, with two primary methods: context-adaptive variable-length coding (CAVLC) and context-adaptive binary arithmetic coding (CABAC). CAVLC uses predefined code tables that adapt based on neighboring macroblock statistics, assigning shorter codes to frequent symbols like small motion vector differences or trailing ones in coefficient levels, while CABAC binarizes syntax elements into binary strings and applies arithmetic coding with context models for up to 10-15% better compression efficiency over CAVLC. These methods ensure that prediction modes (e.g., intra or inter) and vector components are encoded with minimal bits for typical video content. In the standard, the macroblock syntax begins with the address delta (macroblock_address_increment), followed by motion vectors if applicable, and then the 8x8 DCT coefficients scanned in order to group low-frequency components first for . This order facilitates efficient of the DC and AC coefficients, starting from the top-left (DC) and traversing diagonally to prioritize energy compaction. To handle rare events such as unusually large motion vectors or levels beyond standard codeword ranges, escape codes are incorporated into the variable-length coding schemes, allowing explicit representation of these outliers while maintaining compactness for common cases in natural video sequences. This mechanism, present in both MPEG-2's VLC and H.264's CAVLC/CABAC, prevents expansion from infrequent anomalies and supports robust decoding.

Parameter Encoding

In H.264/AVC, motion vectors for inter-predicted macroblocks are encoded differentially relative to predictors derived from neighboring blocks to exploit spatial correlations and reduce bitrate. The predictor is typically the of the motion vectors from the left, upper, and upper-right neighboring blocks (or substitutes if unavailable), and the resulting motion vector difference (MVD) is quantized and entropy-coded using exponential-Golomb variable-length codes that assign shorter bit sequences to small displacements, which predominate in natural video motion. Prediction modes, particularly for intra macroblocks, employ context-adaptive encoding to signal the selected direction efficiently based on adjacent blocks. A key technique is the most probable mode (MPM), which identifies one or more likely modes from the left and upper neighbors; if the actual mode matches an MPM, it is indicated with a single flag (1 bit), otherwise a fixed-length code or enumerated index is used for the remaining possibilities, thereby minimizing bits for common spatial redundancies. Residual data after prediction is transformed, quantized, and represented as a sequence of coefficients that are entropy-coded using to group trailing zeros, paired with level values for non-zero coefficients, and terminated by an signaling the position of the last significant coefficient in the zigzag scan order. Quantization parameter (QP) adjustments specific to each macroblock are encoded via a signed exponential-Golomb-coded delta value (mb_qp_delta), allowing fine-grained control over compression strength without altering the global QP. The macroblock type (mb_type), which defines the overall coding mode (e.g., intra, inter partitions, skip) and associated flags, is encoded using exponential-Golomb codes in H.264/AVC, supporting up to 25 modes in P slices (including intra variants differentiated by coded block patterns), with codeword lengths biased toward frequent types like 16x16 inter partitions to balance rarity and occurrence in typical content. Rate-distortion optimization (RDO) during encoding evaluates candidate parameter sets—such as motion vectors, modes, and QP deltas—by minimizing a Lagrangian cost function combining and estimated bitrate, thereby influencing selections for efficiency, while the standardized bitstream syntax independently ensures deterministic decodability at the decoder.

Artifacts and Limitations

Macroblocking Phenomenon

Macroblocking is a prominent visual artifact in block-based video compression, characterized by noticeable discontinuities that form a grid-like pattern across the image, resembling tiled squares. This phenomenon stems from the independent processing of macroblocks, where each 16×16 pixel unit undergoes separate motion prediction and quantization, resulting in abrupt transitions at their boundaries that do not align with the original continuous scene. Such mismatches are exacerbated by the (DCT) quantization applied within each block, which discards high-frequency details unevenly and introduces artificial edges, especially in regions with low spatial variation. The artifact typically appears as "blockiness" or "tiling," most evident in flat, uniform areas like solid skies or shadowed regions, where subtle gradients are replaced by harsh, squared-off demarcations. At low bitrates, corresponding to high compression ratios, these effects intensify as coarser quantization levels amplify the boundary discrepancies, reducing overall smoothness. Transmission errors, common in error-prone environments such as channels, further aggravate macroblocking by corrupting macroblock data, leading to incomplete decoding and intensified block visibility. In video encoded at typical DVD bitrates around 5 Mbps, this artifact frequently emerges in low-contrast scenes like or clear skies, highlighting the limitations of early block-based codecs under bandwidth constraints. Quantification of macroblocking often relies on objective metrics like (PSNR) and Structural Similarity Index (SSIM), which capture the degradation in fidelity and perceptual structure. Blocking significantly impairs quality when PSNR drops below 30 dB, a threshold where artifacts dominate and viewer satisfaction declines sharply, as coarser quantization preserves less detail across block edges. The term "macroblocking" originated in the amid the development of standards like and MPEG-1/2, which popularized macroblock-based encoding and first documented this boundary-induced distortion in technical literature.

Mitigation Approaches

To mitigate macroblocking artifacts arising from block-based quantization in video compression, codec advancements incorporate in-loop filters that smooth discontinuities at block boundaries during decoding. In the H.264/AVC standard, the adaptive operates within the coding loop to attenuate blocking effects by selectively averaging samples across edges. This filter first determines a boundary strength (Bs) based on prediction modes and motion vector differences between adjacent blocks, then applies thresholds β (for large discontinuities) and t_C (for small changes), both derived from lookup tables indexed by the quantization parameter (QP). Filtering involves low-pass operations, such as averaging four samples with clipping to avoid excessive blurring in detailed areas, thereby preserving sharpness while reducing visible seams. Subsequent standards like HEVC (H.265) build on this by introducing smaller, adaptive block sizes through a partitioning structure, where coding tree units (CTUs) up to 64×64 pixels can be recursively split into coding units (CUs) as small as 8×8, minimizing the perceptual impact of boundaries in uniform regions. Content-aware partitioning further avoids unnecessary splits in textured or complex areas by aligning blocks with image features, such as edges or motion, which reduces quantization mismatches and visible blocking compared to H.264's fixed 16×16 macroblocks. HEVC also enhances deblocking with classification-based decisions that adjust filter strength based on local gradients and QP variations. Post-processing techniques complement in-loop methods by applying non-reencoding filters to decoded output, blending blocks without altering the . Traditional approaches use low-pass filters across suspected boundaries to attenuate high-frequency discontinuities, often guided by to target artifact-prone areas selectively. More recent advancements employ AI-based upscaling in media players, where convolutional neural networks (CNNs) learn to inpaint and denoise blocky regions from training on paired low- and high-quality datasets, effectively restoring detail in real-time playback. The H.264 deblocking filter specifically applies offsets β and t_C from QP-derived tables, clipping pixel modifications to ±t_C pixels to balance smoothing and over-filtering prevention; simulations show it yields average PSNR gains of 0.25–0.35 dB and up to 9–10% bitrate savings at equivalent quality by curbing blocking propagation in prediction references. Looking to future trends, the (finalized in 2018) integrates a switchable loop restoration filter—using Wiener or self-guided restoration on units up to 256×256 pixels—to recover high-frequency details lost to quantization, further diminishing artifacts in real-time decoding scenarios. Emerging integrations of models, such as post-filters tailored for , enable predictive correction of residual blocking during playback, enhancing efficiency for streaming applications.

References

  1. https://www.sciencedirect.com/topics/computer-science/macroblock
  2. https://www.thebroadcastbridge.com/content/entry/19304/compression-part-7-macro-blocks
  3. https://www.vcodex.com/h264avc-inter-prediction
  4. https://iphome.hhi.de/wiegand/assets/pdfs/DIC_video_coding_standards_07.pdf
  5. https://engineering.purdue.edu/~ace/thesis/igarta/thesis-igarta.pdf
  6. https://www.cs.ubc.ca/~krasic/cpsc538a/papers/h264avc-overview.pdf
  7. https://www.boxcast.com/blog/hevc-h.265-vs.-h.264-avc-whats-the-difference
  8. https://www.researchgate.net/publication/228745838_A_Brief_History_of_Video_Coding
  9. https://www.itu.int/ITU-T/worksem/vica/docs/presentations/S0_P2_Sullivan.pdf
  10. https://www.loc.gov/preservation/digital/formats/fdd/fdd000530.shtml
  11. https://www.itu.int/rec/t-rec-h.265
  12. https://www.itu.int/rec/T-REC-H.264/en
  13. https://patents.google.com/patent/US7822123B2/en
  14. https://www.itu.int/rec/T-REC-H.265/en
  15. https://www.maths.univ-evry.fr/pages_perso/vtorri/files/UV/h264/Overview_of_H.264.pdf
  16. https://www.vcodex.com/h264avc-intra-prediction
  17. https://tech.ebu.ch/docs/techreview/trev_293-schaefer.pdf
  18. https://www.itu.int/rec/T-REC-H.264
  19. https://gwliang.com/en/posts/intra-prediction-and-inter-prediction/
  20. https://engineering.purdue.edu/~ace/mobility/mpeg/00618010.pdf
  21. https://www.itu.int/wftp3/av-arch/jvt-site/2003_05_Geneva/JVT-H031r2.doc
  22. https://www.cs.columbia.edu/~delbert/docs/Dueck%20--%20MPEG-2%20Video%20Transcoding.pdf
  23. https://iphome.hhi.de/marpe/download/cabac_ieee03.pdf
  24. https://www.spiedigitallibrary.org/journals/optical-engineering/volume-47/issue-12/120501/Joint-motion-vector-encoding-scheme-with-a-pooled-macroblock-type/10.1117/1.3046713.pdf
  25. https://www.fastvdo.com/spie04/spie04-h264OverviewPaper.pdf
  26. https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/el.2018.7452
  27. https://ocw.unican.es/pluginfile.php/2825/course/section/2777/FinalDraftH.264.pdf
  28. https://ieeexplore.ieee.org/document/4217290/
  29. https://live.ece.utexas.edu/publications/2002/sliu_csvt2002_dctblind.pdf
  30. https://ece.uwaterloo.ca/~z70wang/publications/iciar20_macroblk.pdf
  31. https://www.lifewire.com/macroblocking-and-pixelation-1847333
  32. https://pmc.ncbi.nlm.nih.gov/articles/PMC7113072/
  33. https://www.thebroadcastbridge.com/content/entry/19959/compression-part-12-the-evolution-of-video-compression
  34. https://mcube.lab.nycu.edu.tw/wiki/core/uploads/Research/SelectedTopics/Adaptive%20Deblocking%20filter.pdf
  35. https://ieeexplore.ieee.org/document/6316136
  36. https://www.researchgate.net/publication/255568021_Block_Partitioning_Structure_in_the_HEVC_Standard
  37. https://ieeexplore.ieee.org/document/8939464
  38. https://ieeexplore.ieee.org/document/4657879
  39. https://arxiv.org/pdf/2008.06091
  40. https://www.mdpi.com/2227-7390/12/18/2874
Add your contribution
Related Hubs
Contribute something
User Avatar
No comments yet.