Wavelet Pair Noise Removal for Increasing the Classification Accuracy of a Remotely Sensed Image
- H. S. Jin, H. Y. Yoo, J. Y. Eom, I. S. Choi, D. Y. Han
- Korean Journal of Remote Sensing, Vol.25, No.3…
A method to find the nearly optimal PSNR values for compression was tried to remotely sensed images. The multiresolution method using wavelet transform divide the image to four frequency band, i.e. LL LH HL HH., It analyzes the LL part again as the level goes up. The energy distribution pattern to four parts depends on the kind of wavelet . The energy concentration of LL part was expressed by PSNR value . For large PSNR value there is much energy in LL. Hence the PSNR value can be an index to the multiresolution image compression. In this paper for many wavelet pairs the variation of the PSNR values was analyzed. To calculate the real maximum PSNR value for a given set of wavelets costs too much at the level 3. An algorithm is introduced to get the PSNR value close to maximum one at the very little cost. It is a problem to choose the right wavelet filters. The choice of wavelet has some criterions such as vanishing moment, regularity, and size of support [1, 2]. However, it is hard to get the right wavelet in the real processing. The proposed algorithm does not show the maximum PSNR values for any case but give us a fairly close to the maximum.