Friday, August 28, 2009

IMAGE COMPRESSION

INTRODUCTION:

Digital image compression is a very popular research topic in the field of multimedia processing. Its goal is to store an image in a more compact form, i.e., a representation that requires fewer bits than the original image. It relies on the fact that image information, by its very nature, is not random but exhibits order and has some form of structure. If this order and structure can be extracted, the essence of the information often can be represented and transmitted using less data bits than would be needed for the original. We can then reconstruct the original or a close approximation of it at the receiving end.

Image, video and audio signals can be compressed due to the following reasons:

• Within a single image or a single video frame, there exists significant correlation or redundancy among neighboring samples or pixels. This correlation is referred as spatial correlation or redundancy.

• For data acquired from multiple sensors (such as satellite images), there exists significant correlation or redundancy among samples from these sensors. This correlation or redundancy is called spectral correlation or redundancy.

• For temporal data (such as video sequence), there is significant correlation or redundancy among pixels of successive video frames. This is referred to as temporal correlation or redundancy.


A systematic view of the compression process is depicted in Figure 2.1.



As depicted in Figure 2.1, the source coder performs the compression process by reducing the input image data size to a level that can be supported by the storage or transmission channel. The output bit rate of the encoder is measured in bits per sample or bits per pixel. For image or video data, a pixel is the basic element therefore bits per sample (bps) also referred to as bits per pixel (bpp). The channel coder translates the compressed bit-stream into a signal suitable for either storage or transmission using various methods such as variable length coding, Huffman coding or Arithmetic coding.
In order to reconstruct the image or video data the process is reversed at the decoder. In compression systems, the term ‘compression ratio’ is used to characterize the compression capability of the system.

Compression ratio = Source coder input data size
Source coder output data size

For a still image, size could to the bits needed to represent the entire image. For video, size could refer to the bits needed to represent one frame of video, i.e., one second of video.



4.2 Classifying Compression Schemes

The classification of compression schemes can be done in the following manner.
(a) Lossless vs. Lossy compression: In lossless compression schemes the reconstructed image, after compression, is digitally identical to the original image. However, lossless compression can only achieve a modest amount of compression. On the other hand, lossy schemes are capable of achieving much higher compression but under normal viewing conditions no visible loss is perceived (visually lossless). Some of the lossy compression schemes used include differential pulse code modulation (DPCM), pulse code modulation (PCM), vector quantization (VQ), Transform and Subband coding. An image reconstructed following a lossy compression contains degradation relative to the original. Often this is because the compression scheme also discards non-redundant information.
(b) Predictive vs. Transform coding: In predictive coding, information already sent or available is used to predict other values, and the difference is coded. Since this is done in the image or spatial domain, it is relatively simple to implement and is readily adapted to local image characteristics. The DPCM is one particular example of predictive coding. Transform coding, on the other hand, first transforms the image from its spatial domain representation to a different type of representation using some well-known transforms such as DCT, DWT or Lapped transform, and then codes the transformed values (coefficients). This method provides greater data compression compared to predictive methods as transforms use energy compaction properties to pack an entire image or a video frame into fewer transform coefficients. Most of these coefficients become insignificant after applying quantization, which means less data to be transmitted. In predictive coding, the differences between the original image or video frame samples and the predicted ones remain significant even after applying quantization. This means more data to be transmitted compared to transform coding.


(c) Subband Coding: The fundamental concept behind subband coding is to split the frequency band of a signal (image in our case) in various subbands. To code each subband, we use a coder and bit rate accurately matched to the statistics of the subband.

2 comments:

betitveno said...

nice information
it really helps me
hay thanks for sharing it

Anonymous said...

Nice pretty! http://www.itsolusenz.com/business-model/product-co-development.aspx