Compression Techniques
This research describes an image representation technique thatentails progressive refinement of user specified regions ofinterest (ROI) of large images. Progressive refinement oforiginal quality can be accomplished in theory. However, due toheavy burden on storage resources for our applications, werestrict the refinement to about 25% of the original dataresolution. Wavelet decomposition with Vector Quantization (VQ)of the high frequency components and JPEG/DCT compression of lowfrequency component is used as representation framework. Oursoftware will reconstruct the region selected by the user fromits wavelet decomposition at desired resolution. Further
IMAGE DECOMPRESSIONThe reconstruction program loads the LL subband by decompressingthe Low-Low wavelet component using JPEG/IDCT. IMAGE COMPRESSIONThe entire architecture of the program is based on objectoriented programming using C++. The reconstruction can be iterated on user's request untilsufficient image quality is displayed for the user to makedecision about placing an order for the image. The scalarquantization is heavier for higher resolutions and reducedsuccessively for low resolution high frequency components. Three codebooks for HH, HL, and LH subbands are to be used, butthe same codebook will be used for different resolutions. Thecoefficients are then further compressed by a variant of VectorQuantization (VQ) called Model-Based VQ (MVQ) [3]. The given image is decomposed into low andhigh frequency bands along the rows and columns. We superimpose a HVS modelto convert i. d)random numbers generated from a Laplacian distribution withparameters determined from the respective subbands (HH, HL, LH). Athree level block diagram of reconstruction is shown in Figure6. t from the first preview can be obtained progressivelyby transmitting high frequency coefficients from low resolutionto high resolution, which are compressed by variant of VectorQuantization called Model Based Vector Quantization. Image decomposition is donerecursively into wavelet coefficients using the technique [2]shown in Figure 1. VECTOR QUANTIZATIONThe high frequency subbands are first scalar quantized and thelow frequency band, LL subband of the highest-leveldecomposition is compressed using JPEG/DCT technique.
Common topics in this essay:
HL LH,
IMAGE DECOMPRESSION,
IMAGE COMPRESSION,
Quantization VQ,
,
VECTOR QUANTIZATION,
Vector Quantization,
System HVS,
VQ MVQ,
vector quantization,
HH HL,
hh hl,
hl lh,
hh hl lh,
low frequency,
progressive refinement,
frequency components,
wavelet coefficients,
variant vector quantization,
compressed variant,
low resolution,
variant vector,
compressed variant vector,
coefficients hh hl,
vector quantization vq,
|
Acceptance Essays
