IMAGE COMPRESSION

A presentation

Guided by: Prof. N. A. NEMADE

by: Manish kumar Abhishek ranjan Prashant wagh

Why image compression?

The image compression are required for providing the proper image Format for different devices or application which can be distinguished On the basis of their platform and rendering type.

For Web based applicationsProvides Different image formatsProvides basics for video compression

IMAGE COMPRESSION

GIF (GRAPHICS INTERCHANGE FORMAT)

PNG (PORTABLE NETWORK GRAPHICS

MNG (MULTIPLE-IMAGE NETWORK GRAPHICS)

JPEG (JOIN PICTURE EXPERT GROUP

GIF (GRAPHICS INTERCHANGE FORMAT)

Introduced by CompuServe in 1987 (GIF87a)

Support for multiple images in one file and metadata adding in 1989 (GIF89a)

Indexed image format: up to 256 colours per image, chosen from a variable palette.

One colour index can indicate transparency.

Uses lossless LZW compression of data bytes.

Optional interlacing capability

GIF Uses GIF became very popular in the early days of the

Web.• Supported by NCSA Mosaic.• Pretty good compression.• Most displays then were indexed rather than

truecolor. Today it’s still good for diagrams, cartoons, and other nonphotographic images.

• Lossless encoding good for sharp edges (doesn’t blur).

PNG (Portable Network Graphics)

Supports truecolor, greyscale, and palette-based (8 bit) colourmaps. Uses DEFLATE algorithm:

• As used in gzip.• LZ77 algorithm with Huffman coding.• Patent free.

Combines this with prediction.• 5 different simple prediction algorithms can be used, chosen on a

per-scanline basis.• Eg: sample-to-left, sample-above, average of s-t-l and s-a, etc.• DEFLATE only compresses the difference between the prediction

and the actual value.

JPEG (Joint Photographic Expert Group)

Unlike GIF graphics, JPEG images are full-color images (24 bit, or "true color").

The JPEG come in two flavours, a Standard JPEG and a Progressive JPEG.

First a 'blurred' image is displayed, then as more information flows in, the clarity increases.

Progressive JPEG images often take longer to load onto the page than standard JPEGs, but they do offer the reader a quicker preview.

JPEG images are used mainly for photographs.

JPEG file format uses lossy compression.

JPEG Convert RGB (24 bit) data to YUV.

Typically YUV 4:2:0 used.

Three “sub-images”, one each for Y, U and V

U and V sub-images half the size in each dimension as Y

Divide each image up into 8x8 tiles.

Convert to frequency space using a two-dimensional DCT

Quantize the frequency space, using more bits for the lowerfrequencies.

Encode the quantized values using Run-length encoding and Huffman coding in a zigzag manner.

JPEG DIAGRAM

File Compression

• 2 Types of File Compression: – Lossy Compression– Lossless Compression

• The objective is to reduce redundancy of the image data in order to be able to store data in an efficient form. reducing the file size

Lossy Compression

• A compression scheme in which certain amounts of data are sacrifices to attain smaller file sizes. It is used when the files do not required extra data.

• User can choose a loss rate or ratio.• JPEG uses lossy compression.

Lossy Compression

Quality level: 90 File size: 10,582 bytes

Quality level: 50 File size: 5,154 bytes

Quality level: 1 File size: 923 bytes

Lossless Compression

• Lossless compression means that even though the file is compressed (enjoys a smaller file size than an uncompressed image), it will not lose any quality.

• A lossless image will contain identical data regardless of whether it is compressed pr uncompressed

• Example of lossless compression included LZW (Lempel-Ziv-Welch) and RLE (Run Length Encoding)

Advantages

• It is fast to compute compared to many other suitable image transformations,

• It usually tends to result in low values for the weights of some of its basis images.

• The image that results from deleting low weight components is generally pleasing to the eye.

conclusion

• While the goal was reached in this particular case, this was not the same for all images tested.

• Using Neural networks it can be used to represent portions of the DWT coefficients and save some space. Neural networks generally work better on smaller datasets, (with two neurons in the input layer).