Scribd Logo
Upload

Welcome to Scribd!

  • 25 views

    ECE 576 - Engineering of Computer Based Systems: JPEG Image Compression

    Uploaded by

    jnax101
    1. JPEG images use color transforms like YCbCr to improve compression by representing color information in a way that concentrates most of the visually important information in one channel (Y). 2. It then applies a discrete cosine transform (DCT) to blocks of pixels to represent the image in the frequency domain, followed by quantization which discards less visually important high frequency data. 3. The quantized DCT coefficients are then encoded using run length encoding followed by Huffman coding to further compress the data, which is written to the JPEG file. Decompression involves reversing these steps to reconstruct the image.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd

    ECE 576 - Engineering of Computer Based Systems: JPEG Image Compression

    Uploaded by

    jnax101
    25 views10 pages
    1. JPEG images use color transforms like YCbCr to improve compression by representing color information in a way that concentrates most of the visually important information in one channel (Y). 2. It then applies a discrete cosine transform (DCT) to blocks of pixels to represent the image in the frequency domain, followed by quantization which discards less visually important high frequency data. 3. The quantized DCT coefficients are then encoded using run length encoding followed by Huffman coding to further compress the data, which is written to the JPEG file. Decompression involves reversing these steps to reconstruct the image.

    Original Description:

    Jpeg 020606

    Original Title

    Jpeg 020606

    Copyright

    © © All Rights Reserved

    Available Formats

    PDF, TXT or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    1. JPEG images use color transforms like YCbCr to improve compression by representing color information in a way that concentrates most of the visually important information in one channel (Y). 2. It then applies a discrete cosine transform (DCT) to blocks of pixels to represent the image in the frequency domain, followed by quantization which discards less visually important high frequency data. 3. The quantized DCT coefficients are then encoded using run length encoding followed by Huffman coding to further compress the data, which is written to the JPEG file. Decompression involves reversing these steps to reconstruct the image.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    25 views10 pages

    ECE 576 - Engineering of Computer Based Systems: JPEG Image Compression

    Uploaded by

    jnax101
    1. JPEG images use color transforms like YCbCr to improve compression by representing color information in a way that concentrates most of the visually important information in one channel (Y). 2. It then applies a discrete cosine transform (DCT) to blocks of pixels to represent the image in the frequency domain, followed by quantization which discards less visually important high frequency data. 3. The quantized DCT coefficients are then encoded using run length encoding followed by Huffman coding to further compress the data, which is written to the JPEG file. Decompression involves reversing these steps to reconstruct the image.

    Copyright:

    © All Rights Reserved
    Download as PDF, TXT or read online from Scribd
    Download as pdf or txt
    of 10

    1

    ECE 576 - Engineering of Computer Based Systems


    J PEG Image Compression
    *Thanks to Greg Stitt for providing/assembling much of the included information.
    2
    J PEG Image Compression
    Color Transform
    J PEG capable of using any color model
    RGB, HSI, CMY, YUV, YCbCr
    RGB, HSI, CMY spread useful visual information across 3 components
    Hard to select information to discard
    In YCbCr, most of useful information is in Y space
    Some of Cb and Cr can be discarded
    J PEG can transform color model to improve compression
    3
    J PEG Image Compression
    RGB vs. YCbCr
    RGB
    R Red
    G Green
    B Blue
    YCbCr
    Y Luma (Brightness)
    Cb Cb Chroma (Blue minus Luma)
    Cr Cr Chroma (Red minus Luma)
    Luma - Amount of light that passes through or is emitted from a
    particular area
    Chroma - Difference between a color and a chosen reference color of the
    same luminous intensity
    4
    J PEG Image Compression
    RGB vs. YCbCr
    RGB
    YCbCr
    5
    J PEG Image Compression
    Down Sampling
    Luminance more important than chrominance
    Store less information for chrominance
    Example: 1000x1000 image
    Store 1000x1000 luminance pixels
    Store 500x500 chrominance pixels
    Each chrominance pixel corresponds to actual 2x2 block
    => Each 2x2 block requires 6 pixels instead of 12
    6
    J PEG Image Compression
    DCT Discrete Cosine Transform
    Convert spatial/time domain to frequency
    domain
    Time domain plots how some value
    changes over time
    Frequency domain decomposes signal from
    time domain into sine and cosine waves
    with various amplitudes
    Spatial frequency
    The frequency over distance (instead of
    time)
    The frequency domain is a space in which
    the value at each image position
    represents the amount that the intensity
    values in the image vary over a specific
    distance
    Any image can be formed by adding
    different spatial frequencies
    7
    J PEG Image Compression
    Frequency Domain
    8
    J PEG Image Compression
    DCT Discrete Cosine Transform
    Output of DCT is array of DCT coefficients representing
    different strengths of each frequency
    For J PEG, DCT is done on 8x8 blocks
    8x8 DCT coefficients
    Top left is low frequencies
    (0,0) is DC value, or average, of all pixels
    Bottom right is strength of cosine wave alternating from
    max to min every pixel
    9
    J PEG Image Compression
    DCT Discrete Cosine Transform
    DCT Equation:
    D[x,y] = original pixel value at location x, y
    F(u,v) = output coefficient at location x, y
    10
    J PEG Image Compression
    DCT Discrete Cosine Transform
    DCT operates on 8x8 blocks,
    u, v, x, and y only range in value from 0 to 7
    Precompute 64 possible cosine values and store in an 8x8 table,
    Rewrite DCT Equation:
    D[x,y] = original pixel value at location x, y
    F(u,v) = output coefficient at location x, y
    11
    J PEG Image Compression
    DCT Discrete Cosine Transform
    Reason for using DCT
    Important to separate frequencies because human eye is more
    sensitive to low frequency information
    Can discard some of high-frequency information
    12
    J PEG Image Compression
    Quantization
    Divides each DCT coefficient by some value to reduce total number of
    bits
    Many of the high frequency coefficients become 0
    Other information is lost from integer division
    This is where the loss is J PEG comes from
    .. .. 12 31
    16 15 59 42 32 12 32 42
    12 94 21 35 23 67 2 49
    19 12 23 12 76 32 74 62
    32 12 53 14 23 45 62 89
    /
    0 0 0 0 0 0 0 0
    0 0 0 0 0 0 0 0
    0 0 0 0 0 0 0 0
    0 0 0 0 0 0 0 0
    0 0 0 0 0 0 0 0
    0 0 0 0 0 0 0 0
    0 0 0 0 1 0 1 1
    0 0 0 0 0 1 1 1
    =
    13
    J PEG Image Compression
    Quantization
    Different Types
    Constant amount
    Varying amount for different frequencies
    Separate values for luminance and chrominance
    Larger amount for chrominance
    Quantization table stored in J PEG file
    Generally a sample table is used
    Different Quality jpegs use different quantization tables
    14
    J PEG Image Compression
    Quantization
    Example quantization tables:
    15
    J PEG Image Compression
    Quantization
    Qual = 100
    Size = 60.8 KB
    Qual = 75
    Size = 12.8 KB
    Qual = 50
    Size = 6.9 KB
    16
    J PEG Image Compression
    Quantization
    Qual = 25
    Size = 2.6 KB
    Qual = 10
    Size = 1.8 KB
    Qual = 1
    Size = 1.6 KB
    17
    J PEG Image Compression
    Encoding
    After quantization, many DCT coefficients are 0, especially for high
    frequencies
    High frequencies appear in bottom right of 8x8 block
    Grouping these 0s together makes compression better
    Zig-zag converts 8x8 block into 16x1, in the following order:
    18
    J PEG Image Compression
    Encoding
    Run length encoding
    For each non-zero DCT coefficient, J PEG records the number of
    zeros that preceded the number, the number of bits needed to
    represent the number's amplitude, and the amplitude itself.
    19
    J PEG Image Compression
    Encoding
    After run length encoding, Huffman encoding further
    compresses the 8x8 block
    After compression, block is written to file along with end
    of block marker
    20
    J PEG Image Decompression
    Basically opposite of encode
    Decompress, Dequantize, Inverse DCT, Up-sampling, color
    transform

    You might also like