Video Image Compression CITT H.261 Video Coding Algorithm.

Video Image Compression CITT H.261 Video Coding Algorithm.
A Deep Dive into Video Compression Techniques​

1. Video Image Compression Ishwar Raj​ 23U03007​ Nitin Shewale ​ 23U03020 Pradeep Singh​ 23U03022​ ​ Shreyash Jadhav ​ 23U03036​ CITT H.261 Video Coding Algorithm. A Deep Dive into Video Compression Techniques​

2. Overview Video compression reduces the size of video files to save storage space and enable faster transmission over networks, all while trying to maintain visual quality. Why It Matters: Crucial for streaming platforms (e.g., YouTube, Netflix) Enables real-time video calls (e.g., Zoom, Skype) Optimizes video delivery over mobile networks and limited bandwidth connections Impact: Makes modern video applications possible Drastically reduces file sizes, often by 90–95% Enables technological innovations like streaming platforms and mobile video Video Image Compression Introduction to Video Compression

3. What is Video Coding? Overview Video coding is the process of compressing video data by reducing redundancy to make storage and transmission more efficient. Main Components: Frame Compression (Spatial Redundancy): Reduces repetition within individual frames using transformations and quantization techniques. Temporal Compression: Uses motion estimation and compensation to minimize redundant information between frames. Goal: Balance compression efficiency with video quality to suit different applications like streaming, conferencing, or broadcasting.

4. ITU-T H.261 Standard Overview H.261 is a pioneering digital video compression standard developed by the ITU, mainly for video conferencing over ISDN networks. Key Facts: Finalized in 1990; work began in the mid-1980s Designed for bitrates that are multiples of 64 kbps Supports two resolutions: QCIF (176×144) and CIF (352×288) Enabled real-time video over limited-bandwidth networks Significance: H.261 laid the foundation for modern video codecs and was the first to be widely adopted for practical use in telecommunications.

5. Key Features of H.261 Technical Features: Hybrid Coding Architecture : Combines transform coding with motion compensation Hierarchical Structure : Video organized into pictures, groups of blocks (GOBs), and macroblocks Variable Bit Rate : Adjustable quantization to control output bit rate Loop Filter : Optional spatial filter to reduce blocking artifacts Forward Error Correction : Capability to handle transmission errors Bitrate Control: Target bitrate achieved through adaptive quantization Buffer control mechanisms to maintain consistent transmission rate and Quantization parameter adjusted dynamically based on buffer fullness Color Sampling: YCbCr 4:2:0 color space representation- Full resolution luminance (Y) channel- Half-resolution chrominance ( Cb , Cr) channels horizontally and vertically I-frames (Intra-coded) P-frames (Predictive-coded) Self-contained reference frames Encoded using motion vectors and references to previous frames Picture Types

6. Video Compression Process in H.261 H.261 uses both intra-frame (I-frame) and inter-frame (P-frame) compression. Main Steps: Motion Compensation – Tracks object movement to reduce redundancy between frames. DCT Transformation – Converts pixel data into frequency components for better compression. Quantization – Reduces the precision of DCT output to save space. Entropy Coding (Huffman) – Further compresses data using variable-length codes. Key Techniques: I-Frame Each frame is divided into macroblocks (16×16), further split into smaller 8×8 blocks. DCT is applied to each block to transform spatial data into frequency components. Quantized DCT coefficients are then entropy coded without reference to any other frame. P-Frame Uses motion estimation to find matching macroblocks in the previous frame. Motion vectors are used to predict the current block (motion compensation), and only the residual difference is encoded. The residual undergoes DCT, quantization, and entropy coding like I-frames, but requires fewer bits due to reduced data.

7. Block Diagram of H.261 Encoder The H.261 encoder follows a structured pipeline to compress video: Block Explanations: Motion Estimation & Compensation: Used only in P-frames to predict motion and create residuals. DCT: Converts each 8×8 block to the frequency domain. Quantization: Simplifies DCT output, trading some quality for smaller size. Huffman Coding: Encodes data using shorter codes for common values, improving compression efficiency. Buffer Management: Ensures consistent data rate during transmission.

8. Motion Compensation in H.261 Overview: Reduce temporal redundancy by predicting motion between frames. Block-Matching Algorithm: Each 16×16 macroblock in the current frame is matched to the best-fit block in the previous frame within a ±15 pixel search window. Best match is determined using the Sum of Absolute Differences (SAD). The match is encoded as a motion vector indicating displacement in x and y directions. Motion Vector Details: Represents the horizontal and vertical shift from the reference frame. Differential encoding is used vectors are got from neighboring blocks to save bits. Zero vector used for static areas; poorly matched blocks may be intra-coded. Example: A car moves 8 pixels right and 2 pixels down between two frames. Motion vector: (8, 2), indicating movement from previous to current position.

9. Discrete Cosine Transform (DCT) DCT transforms an 8×8 block of pixel values into frequency components to compress spatial information. It helps separate image content into parts of differing importance (low vs high frequency). Key Properties: Energy Compaction: Low-frequency values retain most visual content. De-correlation: Converts spatial redundancy into independent values. Separable: Can be computed as two 1D DCTs – first rows, then columns. Example: Fig. (a) : Original 8×8 Block (partial) Fig. (b) : After DCT (rounded) 0,0) = DC coefficient → avg brightness Other = AC coefficients → image details Most values near bottom-right ≈ 0 → negligible → compressible [52, 55, 61, 66, 70, 61, 64, 73], [63, 59, 55, 90, 109, 85, 69, 72], ... fig. (a) [665, -21, -19, 9, 17, -9, -3, 1], [-57, -40, -15, 17, 11, 7, -2, 0], ... fig. (b)

10. Quantization in H.261 Quantization is the step where real compression happens by reducing precision of DCT coefficients. How It Works: Each DCT coefficient is divided by a quantization value and rounded. FQ(u,v) = round(F(u,v) / Q(u,v)) Example: If a DCT coefficient is 66 and QP = 10: Quantized value = round(66 / (10×2)) = round(66 / 20) = 3 In H.261: Uses a uniform quantization matrix with a parameter QP (1–31). DC coefficient (top-left) is divided by 8. Other coefficients: divided by QP × 2. Higher QP = stronger compression, but lower quality. increase QP decrease QP (reduce size) (improve quality) If buffer is full If buffer is empty

11. Huffman Coding in H.261 Huffman coding offers lossless compression by assigning shorter codes to more frequent values in the quantized data. Process Overview: Zigzag scan: Converts 8×8 block to 1D sequence. Run-length encoding: Groups zeros and values as (run, level) pairs. EOB (End of Block): Marks end of non-zero values. Huffman codes: Applied to each (run, level) using predefined VLC tables. Example: Quantized DCT: Run-length: (0,12), (0,5), (0,3), (2,1), (3,2), EOB VLC: (0,12) → "011", (3,2) → "00001001011", etc. Final bitstream: Concatenated Huffman codes In H.261: Separate VLC tables for DC, AC coefficients, Motion vectors & Macroblock patterns. 12, 5, 3, 0, 0, 1, 0, 0, 0, 2, 0... Efficiency: Achieves significant size reduction Close to theoretical minimum ( entropy-based )

12. H.261 Decoder Overview The H.261 decoder essentially reverses the encoding process to reconstruct video frames. Key Steps: Bitstream Parsing: Extracts frame, GOB, and macroblock headers. Variable Length Decoding: Huffman decoding → Run-length decoding → Quantized DCT matrix. Inverse Quantization: Multiply by QP: F'( u,v ) = FQ( u,v ) × Q( u,v ) Inverse DCT (IDCT):Converts frequency domain back to spatial pixels. Motion Compensation (P-frames): Applies motion vectors on reference frames. Adds residual for final macroblock. Intra-coded blocks decoded directly. Post-Processing: Optional filtering, color conversion ( YCbCr → RGB), and frame storage. Optimization Focus: Fast IDCT, efficient memory/buffer management Error Handling: GOB resync points Basic error concealment (non-standardized)

13. Applications of H.261 Video Conferencing Systems: Used in H.320 standard systems Supported both desktop and room setups Common in PictureTel , VTEL, CLI Typically ran on ≤384 kbps ISDN lines ISDN-Based Use Cases: Video calls over BRI (128 kbps) & PRI (1.5 Mbps) Remote monitoring, telemedicine, distance learning Multimedia Systems: Early video mail, video phones, answering machines Integrated in business communication tools Notable Products: Intel ProShare , IBM PicTel , CLI Eclipse, AT&T VideoPhone 2500

14. Advantages of H.261 Low Bitrate Support: Worked at 64–384 kbps Acceptable quality at low ISDN speeds Smooth quality degradation & solid rate control Standardized & Supported: Clear specs, reference models Interoperable across vendors Widely adopted under H.32x suite Efficient Implementation: Simple algorithms, no floating-point needed Hardware-friendly design with low memory use Minimal motion estimation effort Real-Time Friendly: Low latency, fast processing Predictable performance, no B-frames Simple buffer handling

15. Limitations of H.261 Compression Efficiency: No B-frames or sub-pixel motion Basic quantization & entropy coding Few intra-prediction modes ~2–3x less efficient than H.263, ~5–10x than H.264 Technical Gaps: No advanced error handling No scalability, region-based coding, or perceptual tuning Format & Resolution Limits: Max resolution: CIF (352×288) 4:2:0 only, no interlacing or custom sizes Frame rate capped at ~30 fps No aspect ratio control Visual Quality Issues: Heavy blocking artifacts at low bitrates Weak texture & detail retention Poor motion handling, minimal filtering

16. Motion Compensation - Solved Example Scenario: A ball moves from (100,150) → (108,152) in two video frames. Find motion vector for a 16×16 block & Compute prediction error (residual) Steps: Compare 16×16 macroblocks in a ±15 pixel search window using SAD. (8,2): SAD = 325 (7,3): SAD = 412 (8,3): SAD = 380 Block Matching Sample SADs Motion Vector (8,2) → Best match (min SAD) Residual Difference between actual and predicted block Efficiency Residuals are smaller → Less data after compression (DCT + quantization)

17. DCT and Quantization - Solved Example Input: An 8×8 pixel block from an image. Steps: Converts pixel values to frequency domain Top-left = brightness Edges = high-frequency details DCT Visuals: Quantization (QP = 10) DC step = 8, others = 20 Most high-frequency values become 0 Compression Original: 64 bytes After quantization: Many zeros Zigzag + RLE = ~4–8× compression Reconstruction Inverse DCT gives slightly lossy but close result Shows efficiency of lossy compression

18. Modern Standards - Summary Codec Year Block Size Motion Bitrate ↓ Notes H.261 1990 16×16 Integer - First standard H.263 1996 16×16 Half-pel 30–50% Improved prediction H.264 2003 16×16 to 4×4 ¼-pel 40–50% HD, streaming H.265 2013 64×64 to 4×4 ¼-pel 50% 4K, HDR, VR ready Codec Year Block Size Motion Bitrate ↓ Notes

19. Future Trends in Compression Next-Gen Codecs: VVC (H.266): 50% better than H.265 AV1: Royalty-free HEVC alternative LCEVC: Adds layers to existing codecs AI-Based Compression: Neural networks for prediction, transforms Up to 30% better than HEVC Emerging Research: Point cloud, light field, low-latency video, energy-efficient encoding Big challenge: Balancing quality, speed, and power Cloud-Based Solutions: Smart encoding using powerful servers Real-time bitrate adaptation (e.g., Netflix, YouTube)

20. Conclusion H.261 was the first practical video compression standard, setting the stage for all modern video codecs. Introduced key concepts like motion compensation, DCT, and intra/inter-frame compression. Enabled early video conferencing over limited bandwidth (ISDN). Though outdated today, it laid the foundation for successors like H.263, H.264, and H.265. Understanding H.261 helps us appreciate how modern video compression evolved.

21. Thank You Conclusion H.261 was the first practical video compression standard, setting the stage for all modern video codecs. Introduced key concepts like motion compensation, DCT, and intra/inter-frame compression. Enabled early video conferencing over limited bandwidth (ISDN). Though outdated today, it laid the foundation for successors like H.263, H.264, and H.265. Understanding H.261 helps us appreciate how modern video compression evolved.