You are on page 1of 4

The basic idea of block transform coding is to transform the image from the spatial domain to another domain

where coefficients have lower entropy so they can be coded more efficiently This approach is used in both JPEG and MPEG because it yields much higher Compression rates than previous methods. We subdivide the image into nxn blocks prior to calculating the transform to improve speed and also to localize any errors that may occur

Transform Selection:
Vary the subimage size from 2x2, 4x4, 8x8, Perform transform Discard smallest P% of T(u,v) coefficients Perform inverse transform Calculate rms error

Coefficient Quantization:
The process of discarding the P% smallest T(u,v) coefficients is a form of quantization In general, quantization will reduce the range of T(u,v) values, which reduces entropy and increases the compression rate A wide variety of quantization schemes have been devised over the years In some systems, the nxn mask for each block is transmitted as an n2 bit stream, followed by the non-zero T(u,v) values for the block Other systems save bits by using a single mask for the entire image, or by predefining a set of common masks, and transmitting the mask identifiers instead of the masks

Coefficient Coding
T (u,v) coefficients are often reordered using a zig-zag pattern prior to coding Low frequency terms before high frequency terms A special symbol EOB is used to replace trailing 0s. After quantization, the T(u,v) values for each block are encoded using a lossless coding scheme such as Huffman or Arithmetic coding The specific parameters of the coding scheme are stored in the compressed file, so the decompression program can undo the coding, rescale the data, and perform the inverse transform to obtain the output image We describe the design of the feature descriptor codec. Since image features are unordered and distinct from each other, we have designed our codec to operate on each descriptor individually. Independent encoding gives the advantage of having random access to a given feature without the need to reference other features. The goal of our codec is to compress each feature individually by using intrinsic descriptor statistics. Digital Image Processing allows you to apply powerful transform domain techniques to your data. Transform methods are typically used in filtering, compression, and image texture analysis. Transform coding compresses image data by representing the original signal with a small number of transform coefficients. It exploits the fact that for typical images a large amount of signal energy is concentrated in a small number of coefficients.

Transform coding is an integral part of the Joint Photographic Experts Group (JPEG) standard for lossy image compression. This loads one of the example images.

Block transform coding divides an image into blocks of equal size and processes each block independently. Block processing allows the coder to adapt to local image statistics, exploit the correlation present among neighboring image pixels, and reduce computational and storage requirements. The baseline JPEG algorithm uses nonoverlapping blocks of dimensions 8x8. Here we compute the block discrete cosine transform of an example image. As required by JPEG, the image is first normalized with respect to the middle gray level.

Here a multiblock segment of the original image and the corresponding DCT coefficient blocks are shown.

You might also like