TRANSFORM

CODING
 Transform coding
Transform coding compresses image data by representing the original
signal with a small number of transform coefficients.
 Sub Image Selection

• At this stage image is divided into a set of sub-images.

• N*N image is decomposed into set of sub-images of size n*n.
• The values of n is a power of 2.
• This process is done to ensure that correlation among the pixel is
minimum.
• This step is necessary to reduce computational complexity.
 Transform Selection
• Idea behind transform coding is to use mathematical transforms for data
compression.
• Transforms like DFT,DCT and wavelet transforms are used.

• Choice of transforms depends on the resources and amount of error

associated with reconstruction process.
• Mathematical transforms are tools for information packing.

• DCT offers better information packing capacity.

• KL transforms are also effective but the disadvantage is that they are data
dependent.
• Digital cosine transform is preferred because it is faster so can pack more
information.
 Bit Allocation
• Transform coding is the process of truncating, quantizing and coding
the coefficients of the transformed sub-image.
• Bits must be assignd so that compressed image have minimum distortion.
• Bits allocation must be done on the importance of data.

Steps involved in bit allocation are:

1. Assign predefined bits to all classes of data in the image.
2. Reduce the number of bits by one and calculate the distortion.
3. Identify the data which is associated with minimum distortion and reduce
one bit from its quota.
4. Find the distortion rate again.
5. Compare the target and if required repeat steps 1-4 to get the optimal
value.
 Zonal coding
This process involves multiplying each transform coefficeient by the
corresponding element in zonal mask, which has 1 in the locations of
maximum variance and 0 in the other places.
• A zonal mask is designed as a part of this process.
• These coefficients are retained because they convey more image
information.
• The locations are identified based on the image models used for source
symbols encoding.
• The retained coefficients are quantizd and coded.
• Number of bits allocated may be fixed or may vary based on some optimal
quantizer.
M(u,v)= 1 if T(u,v) has large coefficients
0 otherwise
 Threshold mask
Threshold coding works based on the fact that transform coefficients having
the maximum coefficients having the maximum magnitude make the most
contribution to the image.

The threshold may be one of the following:

• Single global threshold
• Adaptive threshold for each sub-image
• Variable threshold as a function of the location for each co-efficient in the
sub-image.
The thresholding and quantization process can be combined; their
approximation is
T ^ (u , v)  Round(t (u , v) / z (u , v))

T^ (u,v)- Thresholded and quantized value

T(u,v)- Original Transform co-efficient
z(u,v)- Transformed Normalized array

At the decoded end, T(u,v) = T^(u,v) * Z(u,v) is used to denormalize the

transform co-efficients before inverse transformations
THANK YOU