Digital Image Compression and JPEG Images - lect 8

JPEG Images

• JPEG is a prevalent image format.

• Most digital cameras save images in JPEG format.

• There are two main JPEG image formats, the original and a more recent one based on wavelets (which won't be covered).

Importance of Understanding Image Compression

• Forensic image analysts emphasize the importance of understanding image compression and JPEG for presenting images in court.

• Saving an image as JPEG results in loss of information, making it a lossy compression format.

• However, the loss of information is usually small.

Why Compress Digital Images?

• To reduce storage requirements.

• To decrease bandwidth usage, enabling faster loading of images on websites.

Compression Ratio

• Controlling the amount of compression allows for a trade-off between image quality and file size.

• High compression ratios lead to distortion and loss of detail.

• Lower compression ratios result in less distortion but require more storage.

• It's a compromise between image quality and file size.

Storage Requirements for Uncompressed Images

• Image dimensions: $M$ (rows) by $N$ (columns)

• Number of color channels: $n$ (typically 3 for red, green, and blue)

• Bits per pixel per color channel: $m$ (typically 4 or 8 bits)

• Total storage requirements for uncompressed image: $M \,\times N \,\times m \,\times n$

• Bitmap format is an example of an uncompressed image format.

Lossless vs. Lossy Compression

• Non-lossy/lossless compression: The original image can be recovered exactly from the compressed file (e.g., PNG format).

• Lossy compression: Information is lost during compression, resulting in an approximation of the original image (e.g., JPEG format).

• Lossy compression achieves a much higher degree of compression compared to lossless compression.

• PNG is suitable for retaining all image details, while JPEG is better for minimizing file size and bandwidth requirements.

Data Redundancy

• All image compression schemes utilize data redundancy. There are 3 types:

  1. Coding redundancy.

  2. Interpixel redundancy.

  3. Psychovisual redundancy.

Coding Redundancy

• Caused by suboptimal code words for symbol encoding.

• A symbol is typically a gray level within the image.

• In natural images, intensity values do not occur with equal probability.

• Assign shorter binary strings to frequently occurring intensity values to reduce storage requirements.

• Huffman encoding is an optimal way to achieve this.

• Example: A photograph of a polar bear in a snowstorm will contain mostly white pixels, so we assign a shorter string to encode them.

Interpixel Redundancy

• Encoding the structure efficiently within the image.

• Images have structure that can be utilized to reduce storage requirements.

Psychovisual Redundancy

• Information within an image that is superfluous to interpretation or aesthetics.

• If the eye can't perceive the detail, there is no need to encode it.

• Spatial frequency is the rate of change from light to dark.

• As spatial frequency increases, the ability to perceive differences diminishes.

• As the amplitude of differences decreases, the ability to perceive differences diminishes.

• If fluctuations between light and dark are imperceptible, they don't need to be encoded, allowing for compression by removing that information.

JPEG Images

• JPEG is a prevalent, lossy image format widely used in digital cameras.

• There are two main JPEG formats: original and wavelet-based (not covered here).

Importance of Understanding Image Compression

• Forensic analysts need to understand image compression and JPEG for court.

• JPEG compression is lossy but typically involves small information loss.

Why Compress Digital Images?

• Reduces storage and decreases bandwidth for faster loading.

Compression Ratio

• Balancing image quality and file size involves controlling compression.

• High compression leads to distortion; lower compression requires more storage.

Storage Requirements for Uncompressed Images

• Image dimensions: $M$ (rows) by $N$ (columns)

• Color channels: $n$ (typically 3 for RGB)

• Bits per pixel per channel: $m$ (typically 4 or 8 bits)

• Total storage: $M \times N \times m \times n$

• Bitmap format is uncompressed.

Lossless vs. Lossy Compression

• Lossless: Original image recoverable (e.g., PNG).

• Lossy: Information lost, approximates original (e.g., JPEG).

• Lossy achieves higher compression.

• PNG retains details; JPEG minimizes file size.

Data Redundancy

• Image compression uses data redundancy:

  1. Coding redundancy

  2. Interpixel redundancy

  3. Psychovisual redundancy

Coding Redundancy

• Results from suboptimal code words.

• Assign shorter strings to frequent intensity values (e.g., Huffman encoding).

• Example: Polar bear photo with mostly white pixels.

Interpixel Redundancy

• Efficiently encoding image structure to reduce storage.

Psychovisual Redundancy

• Information superfluous to perception.

• High spatial frequency or low amplitude differences are imperceptible and can be removed to compress.

JPEG Images

• JPEG is a prevalent, lossy image format widely used in digital cameras.

• There are two main JPEG formats: original and wavelet-based (not covered here).

Importance of Understanding Image Compression

• Forensic analysts need to understand image compression and JPEG for court.

• JPEG compression is lossy but typically involves small information loss.

Why Compress Digital Images?

• Reduces storage and decreases bandwidth for faster loading.

Compression Ratio

• Balancing image quality and file size involves controlling compression.

• High compression leads to distortion; lower compression requires more storage.

Storage Requirements for Uncompressed Images

• Image dimensions: $M$ (rows) by $N$ (columns)

• Color channels: $n$ (typically 3 for RGB)

• Bits per pixel per channel: $m$ (typically 4 or 8 bits)

• Total storage: $M \times N \times m \times n$

• Bitmap format is uncompressed.

Lossless vs. Lossy Compression

• Lossless: Original image recoverable (e.g., PNG).

• Lossy: Information lost, approximates original (e.g., JPEG).

• Lossy achieves higher compression.

• PNG retains details; JPEG minimizes file size.

Data Redundancy

• Image compression uses data redundancy:

  1. Coding redundancy

  2. Interpixel redundancy

  3. Psychovisual redundancy

Coding Redundancy

• Results from suboptimal code words.

• Assign shorter strings to frequent intensity values (e.g., Huffman encoding).

• Example: Polar bear photo with mostly white pixels.

Interpixel Redundancy

• Efficiently encoding image structure to reduce storage.

Psychovisual Redundancy

• Information superfluous to perception.

• High spatial frequency or low amplitude differences are imperceptible and can be removed to compress.