Research Of Image Enhancement Based On Multiscale Transform

Content: An image system consists of acquisition, transmission, reception and display, every part of which can be disturbed, so that the image quality can be degraded. Elementary requirement of image processing is how to deal with these degraded images to meet our need. One of the main contents of image processing is image enhancement. Now, there are many methods of image enhancement, but they will more or less bring bad effects when enhancing images. Wavelet analysis, ridgelet analysis and curvelet analysis have predominant advantages in image de-noising and image enhancement, due to their extracting multi-resolution characters of signal and making the difference between noise and signal very clear. Image de-noising and image enhancement based on multiscale transform are studied in the thesis. The main works are as follows:Firstly, the traditional algorithms for image enhancement have the shortcoming of noise over-enhancement. In the thesis, on the method of wavelet image enhancement based on soft threshold we propose a method of wavelet image enhancement based on modulus square processing to achieve a better visual effect. Experiment result shows that this algorithm avoids over-enhancement of noise while enhancing image details with good visual effect.Secondly, a new multiscale transform, called Ridgelet Transform, was proposed by professor Candes,E.J. and Donoho,D.J. in 1999. It is especially suitable for describing the signals, which have linear or super-plane singularities, and have much better approximation. Later, M.N.Do and M.Vetterli proposed an orthonormal version of the ridgelet transform for discrete and finite-size images, named Finite Ridgelet Transform. In order to show the advantage of Finite Ridgelet Transform we apply Finite Ridgelet Transform to image denoising, which is good at image denoising. And we proposed an algorithm of ridgelet image enhancement based on modulus square processing. Experiment result shows that the new algorithm is smoothing gaussian noise, at the same time the algorithm is good at enhancing detail of image.Thirdly, the curvelet is more suitable for image processing than the wavelet, able...