I am looking for a proper and reasonable way to set the number of bases to be used in 2D image compression or denoising. If anyone could give me a hint at least where to look it will be appreciated.
The number of bases depends on the extent or degree to which we want to preserve the image content. An important and popular measure is the PSNR of the image with the compressed representation; if you desire high PSNR you need to choose more bases to meet the requirements. For a given number of Bases, you can have an algorithm to choose the right Bases that give you the maximum PSNR.
Regarding denoisinng, one must know what type of noise you want to remove; we must have the frequency range of the noise to answer this question.
That depends on your image characteristics, usually hte image is divided into tiles of 8x8 pixels then transformed into the DCT domain. Then you apply a Rate - Distortion control by controlling the bit rate. Rate can be controlled by using quantizers.First you need to decide the number of bits to allocate for each band that is, you habe B bits that you have to distribute among M coefficients then decide the stepsize of the quantizer. If you use a quantizer with a large step size a large number of coefficients will be quantized to zero. If you require a higher rate select a quantizer with smaller step size.
For denoising using DCT the best is use a Bayesian method, once you calculate the noise variane you can use a Bayes threshold to denoise the image.
Good luck !
Difficult and impossible to find a global optimal q-table, it is possible to find for specified images
Impossible to set the number of basis to be used. Every image is different.
please take a look at papers O. Pogrebnyak, V. Lukin, Wiener DCT Based Image Filtering, Journal of Electronic Imaging, 2012, No 4, 14 p. ; N. Ponomarenko, S. Krivenko, V. Lukin, K. Egiazarian, Lossy Compression of Noisy Images Based on Visual Quality: A Comprehensive Study, EURASIP Journal on Advances in Signal Processing, 2010, 13 p., doi:10.1155/2010/976436; Lukin V., Ponomarenko N., Egiazarian K., HVS-Metric-Based Performance Analysis Of Image Denoising Algorithms, Proceedings of EUVIP, Paris, France, 2011, pp. 156-161. All are available. If problems write to me to [email protected]
- Badges
- Science topic
- Similar topics
- Phytochemistry
- Extracts