A method for wavelet based image compression by using the Hurst exponent is presented here. Image segmentation is one of the most important steps in image analysis. We decompose an image into two parts that have a strong correlation to objects or areas of the real world. The first part corresponds to edge and trend information. It is obtained either by multiscale edge detection or by thresholding a local Hurst exponent. From this information a partial compressed image is created by thresholding. The second portion of the decomposition contains texture information. It is obtained by taking the difference between the original image and the image constructed from the multiscale edge-trend based encoding. This is referred to as the error image. The texture information, in the form of the error image, is further decomposed by the local Hurst exponent method or rescaled range analysis. The texture parts are then separately compressed by a wavelet or iterated function system (IFS) method.
展开▼
