IEEE Transactions on Pattern Analysis and Machine Intelligence
Image segmentation by histogram thresholding using fuzzy sets
IEEE Transactions on Image Processing
| Hi-index | 0.00 |
Fuzzy methods provide a versatile means in the processing of digitized medical images. Fuzzy thresholding is considered to have certain advantages over crisp thresholding since it allows for better blending of different image segments and facilitates segment dependent weighted application of consequent processing methods. A novel methodology is proposed in the context of this paper. It exploits the morphology of the histogram of digital intra-operative cholangiography images in order to determine the fuzzy membership functions associated with various image segments. Guidelines are provided in order to determine the number of different segments as well as the threshold values between the segments. The method models the gray-level histogram of the images as a mixture of Radial Basis Functions (RBFs), which are consequently used to yield the fuzzy membership functions as kernel regression estimates. The proposed method compares favorably against standard fuzzy thresholding methods that minimize a measure of fuzziness of the histogram of an image like Shannon's entropy and Yager's measure. It allows for histogram compensation of non-uniformly exposed medical images acquired by mobile equipment so that useful diagnostic information may be derived from them.