Based On Local Characteristics Of The Wrist Bone X-ray Image Enhancement Algorithm

The value of the skeletal maturity (i.e. bone age) has been wildly used in the area of prophylaxis, clinical medicine, athletic sports, and justice, etc. It becomes more and more important to these areas. The assessment of skeletal age is drawing data through observing the hand-wrist photography. By the demand of assessment's authenticity and objectivity, many research departments devote to this area, and they have made tremendous achievements.Bone age automatic assessment can be divided into four periods: preprocessing of X-ray hand-wrist image, image contrast enhancement, feature extraction and feature matching. Medical image enhancement is a main subject of modern medical image processing, and it's target is to improve the feasibility of image. In this thesis, algorithm research of image enhancement is mainly focused. Based on X-ray hand-wrist image and non-uniformly illuminated image's characteristics, a new enhancement algorithm is presented.Firstly, this thesis introduce the image enhancement methods of spatial domain, and these methods' application in medical area. Although their results are not so good to hand-wrist image, they provide the theoretic foundation to the localized enhancement. Secondly, the classical localized enhancement algorithm is given. Their theory and process are introduced in detail, and their shortcoming is also analyzed.According to the characteristics of X-ray hand-wrist images, a new enhancement method is presented, which is based on image local identity. This algorithm has introduced some image's texture descriptors. The background and the target region have different intensity mapping model. Different from the classical methods, it brings forward the overlap window's idea. This can make the mapping function become more reasonable, and the practice has proved it. The practical results show that this algorithm can not only enhance X-ray hand-wrist image clearly with less background noise, but also get satisfied results for other non-uniformly illuminated image.