Medical Image Processing Of Chest X-rays

Medical image processing has become an active research area in the community of computer vision. The paper focuses on the research of some key medical image processing technologies in chest X-rays, including chest Radiography images enhancement, segmentation and focus recognition of lung.An improved enhancement algorithm of chest radiographs based on Retinex was presented. The three different Gaussian filter coefficients under three different deviations were calculated, and the convolution operation was implemented between the image distribution and Gaussian filters, a weighted average of multi-scale was gained and mapped to gray range of display device. With this method, image contrast enhancement, sharpening and dynamic range compression were achieved at the same time. The information of hidden area of chest radiographs was obviously enhanced. Improved MSR can overcome the lack of enhancement of traditional methods and satisfy the clinic demand.As for the segmentation of the lungs in chest radiography images, an improved live-wire algorithm for interaction image segmentation is presented. In comparison with the existing algorithm, the proposed one, with the same complexity, greatly improves the performance of image segmentation. Meanwhile, the modified algorithm overcomes the following drawbacks of traditional one: rather sensitive to noise,inefficient to distinguish between strong and weak edges,inapplicable to segmenting images with sharp edges.We introduce Gabor filters in focus recognition of lung, since they have been successfully applied for pattern recognition. An improved Gabor feature extraction algorithm based on feature weighting is proposed. It weights the raw features derived from 2D Gabor filters according to their own degree of dispersion. The experiment results indicate that the proposed method is superior to conventional ones in terms of robustness and discrimination ability. Thus it is fit for the recognition of chest radiography images with poor quality. Possible image locations and sample focus verify those hypothesis using Gabor filters for feature extraction and BP neural network for classification, the focus of lung is recognized at last.