Image compression is very important for their efficient transmission and storage. Images contain large amounts of information that requires much storage space, large transmission bandwidths and long transmission times. Therefore it is advantageous to compress the image by storing only the essential information needed to reconstruct the image. An image can be thought of a matrix of pixels (or intensity) values. Image compression standards bring about many benefits, such as:

• Easier exchange of image files between different devices and applications;
• Reuse of existing hardware and software for a wider array of products;
• Existence of benchmarks and reference data sets for new and alternative development.

Compression is the process of reducing the amount of space required to store the data [6]. Compression is mainly classified into two types namely:

• Lossy compression
• Lossless compression

In Lossy Compression, the compression techniques achieve high compression rates. They reduce the accuracy of the compressed image by producing some distortions. These distortions reduce the quality of image. The Lossless Compression techniques are used in video compression and Real time telecasting. In Lossless Compression, the original image can be reconstructed without loss of data. This kind of compression is called Reversible Technique. The Lossless Compression techniques are used in medical images and satellite images.

The Lossless compression, involves no loss of information. The process of lossless compression basically uses the encoded i.e., compressed image in order to recover the original image. The basic nature of this method is that it is a noiseless and entropy based method because it does not add any sort of noise to image and makes use of methods like statistics/decomposition to exclude the redundancy factor. Lossless compression has limited uses by as in inflexible necessities such as medical imaging. The following are the some of the Lossless Compression techniques:

• Huffman Coding
• Arithmetic Coding
• Run Length Coding
• LZW (Lempel–Ziv–Welch) Coding
• Lossless Predictive Coding

1.1 Huffman Coding [12]

It uses the smallest number of code symbols per source. It is one of the most popular techniques to remove redundancy. The frequency distribution is calculated in order to calculate the probability distribution. The probabilities are higher for shorter code and lower for longer codes. The binary tree is created by using the symbols as leaves according to the probabilities and paths the area taken as code word.

Advantages

• it provides the smallest number of code symbols per source symbol.
• It provides an optimal code for symbols.
• Encoding and Decoding processes are done in a simple lookup table manner.

Disadvantages

• The limit is a maximum of one symbol at a time.
• It is difficult for large number of Images.
• Computational complexity is more.

1.2 Arithmetic Coding [1]

In Arithmetic Coding, there is no one to one correspondence between the source and the code words. The entire sequence of source symbols is assigned to be a single arithmetic word. The code word is defined in the interval of real numbers between 0 and 1. As the number 0 symbols in the message increases, the interval used to represent it becomes smaller thus the number of information bits required to represent the interval becomes larger.

Limitations [6]:

• The end of message indication is required to separate the messages.
• It uses finite arithmetic precision.

Advantages:

• Ability to keep the coding and the modular parts separate.
• No code trees are transmitted to the receiver.
• It uses fractional values.

Disadvantages

• It has complex operation.
• There are no finite precisions.
• Two structures are needed to store the numbers and the constant division of interval result in code overlap.

1.3 Run Length Encoding [11]

It is the simplest technique of all the compression techniques. The consecutive sequence of the symbols are identified as runs and the other are identified as non runs. It checks whether the symbol is repeating or not and is based on redundancies and their lengths. It uses those runs to compress the original source file while keeping all non runs without using the compression process.

Advantages:

• This algorithm is used for small number of inputs.
• It is easy to implement.

Disadvantages:

• It is efficient when the file contains repetitive information.
• Black and white areas are suitable rather than color images.

1.4 LZW Coding [3]

It is a dictionary based, encoding scheme that uses fixed length code words to represent variable length strings. The dictionary is created dynamically in this process and and and there is no need to transfer for decompression process. In decompression the same dictionary is created. It is used for reducing the inter pixel redundancies of the image. This technique is mainly used for in many file formats including PNG, GIF, TIFF and PDF.

Limitations:

• The dictionary needs to be flushed when it becomes full, and a new one should be used to continue with.
• The Least used dictionary entries need to be tracked.

Advantages:

• This is the simplest technique.
• The Dynamic code word is built for each file.
• Decompression creates the code word table so it does not need to be passed.

Disadvantages:

• Actual compression is hard to predict.
• It only works when the input data is sufficiently large.
• It occupies more storage space.

1.5 Lossless Predictive Coding [8]

The decomposition of an image into biplanes or binary is not necessary. The new information is extracted and coded, and then inter pixel the inter pixel redundancies of closely spaced pixels are eliminated. Figure 1 shows the block diagram of lossless predictive coding.

New information = (Actual value) - (Predicted value)

Predictor produces the anticipated value and the value is based on the past inputs. The output of the predictor is rounded to the nearest integer.

Figure 1. Lossless Predictive Encoding

Advantages:

• Large amount of inter pixel redundancies can be eliminated.
• The input image does not need bit plane decomposition.

Disadvantages:

• In video compression, the error may occur in the final frames.

In any compression technique, accuracy is very important in compression and decompression. There will be a possibility of data/information loss, but it should be under the limit of tolerance. It should be good enough for application of image processing. This kind of compression is used for sharing, transmitting or storing multimedia data, where some loss of data or image is allowed [5]. In Lossy Compression, some loss of information is acceptable. In the case of image, there will be some loss of resolution. The following are the some of lossy compression techniques.

• Transform Coding.
• Wavelet Coding.
• Lossy Predictive Coding.

2.1 Transform Coding [2]

Transform Coding technique is based on modifying the transform of an image. The Figure 2 shows the various processes involved in the transformation. A reversible linear transform is used to map the given image into a set of transform coefficients and these coefficients are quantized and coded. The image can be discarded some lower pixels of the image can be discarded and the remaining pixels can be compressed. When the output is decoded, the resulting image is not the same as the given image.

Figure 2. Process in Transformation

Advantages:

• Subimage size is an important factor that reduces the computational cost, and transform error coding.

Disadvantages:

• The reconstruction error in image depends on the precision used to represent the retained values.

2.2 Wavelet Coding [10]

This transform is based on Discrete Wavelet Transform. It decomposes the image into a set of functions called Wavelets. Figure 3 shows the stepwise process of the wavelet encoding system. These wavelets are of limited duration. Initially the transformation begins by transforming the original image into decorrelated image and the important visual information are packed and the remaining small number of coefficients can be truncated or set to zero. This makes the image that has only little distortion.

Figure 3. Wavelet Encoding System

Advantages:

• It reduces the computational efficiency.
• There is no need to subdivide the image.
• The removal of subdivision of images eliminates the blocking artifacts.

Disadvantages:

• The number of decomposition levels affect the reconstruction error and computational complexity.

2.3 Lossy Predictive Coding [7]

It is the Spatial Domain method that operates directly on the image pixels. The Quantizer is used to perform the function of nearest Integer block and it maps the prediction error into the limited range of outputs [9] . Figure 4 shows the encoding formula for a replaced pixel and Figure 5 shows the decoding formula for compressed pixels. The Predictor can predict the value based on the past predictions.

Figure 4. The Encoding Formula for Replaced Pixel

Figure 5. The Decoding Formula for Compressed Pixel

The predictions made by both, encoder and decoder are the same. The predictor is placed within the feedback loop.

Advantages:

• Prediction is based on original unquantized samples.
• Both encoder and decoder offers samples for the prediction.

Disadvantages:

• It is not suitable for lossy image coding at high bit rates.

This paper is a survey of the different compression algorithms. The authors took the survey on both lossy compression and lossless compression techniques. Each compression technique is good for some particular area. Each compression algorithm has advantages and disadvantages. These disadvantages can be overcome in future by using the neural network algorithms with these techniques.