deposit_your_work

Image enhancement by nonlinear wavelet processing

Files in this item

Files Size Format View
Ode1994Oct5Imageenha.PS 9.177Mb application/postscript View/Open

Show full item record

Item Metadata

Title: Image enhancement by nonlinear wavelet processing
Author: Odegard, Jan E.
Type: Conference Paper
Keywords: Temporary
Citation: J. E. Odegard,"Image enhancement by nonlinear wavelet processing," in Wavelets and Large-Scale Image Processing, Int'l. Press, Argonne National Lab,
Abstract: In this paper we describe how the theory of wavelet thresholding introduced by Donoho and Johnstone can successfully be applied to two distinct problems in image processing where traditional linear filtering techniques are insufficient. The first application is related to speckle reduction in coherent imaging systems. We show that the proposed method works well for reducing speckle in SAR images while maintaining bright reflections for subsequent processing and detection. Secondly we apply the wavelet based method for reducing blocking artifacts associated with most DCT based image coders (e.g., most notably the Joint Photographic Experts Group (JPEG) standard at high compression ratios). In particular we demonstrate an algorithm for post-processing decoded images without the need for a novel coder/decoder. By applying this algorithm we are able to obtain perceptually superior images at high compression ratios using the JPEG coding standard. For both applications we have developed methods for estimating the required threshold parameter and we have applied these to large number of images to study the effect of the wavelet thresholding. Our main goal with this paper is to illustrate how the recent theory of wavelet denoising can be applied to a wide range of practical problems which does not necessarily satisfy all the assumptions of the developed theory.
Date Published: 1994-10-20

This item appears in the following Collection(s)

  • ECE Publications [1032 items]
    Publications by Rice University Electrical and Computer Engineering faculty and graduate students
  • DSP Publications [508 items]
    Publications by Rice Faculty and graduate students in digital signal processing.