Adaptive filtering of digital image signals

摘要

A method for adaptive digital image signal filtering for fast magnetic resonance (MR) image processing is provided. A minimum modification to the original image structures is tolerated, while noise smoothing is carried out at a prescribed filtering level. The filtered image is a weighted combination of four sub-images which have been obtained from the low-pass filter of the original image along four main directions. The weighting of each sub-image is controlled by the differences owing to these sub-images and the original image. The resulting image is then focused asymetrically in order to magnify the boundaries of the image structure. The overall effect of this filter structure is an effective adaptive noise reduction and edge magnification with efficient execution accompanied by the use of matrix computers. The high regulatity and the high parallelism of the process render the latter suitable for efficient execution in conjunction with the use of high grade integrated (VLSI, very large scale integrated) circuits or multiprocessor systems. The efficiency of the filter for effective reduction of image noise and the protection of important image structures are discussed and demonstrated with the use of a plurality of MR images from an MR imaging system having a low field strength. <IMAGE>