The Key Image Processing Techniques. Digital Chest

The paper focuses on the research of some key image processing technologies in Digital Chest Radiography Images(abbr. DR),including image segmentation, de-noising and enhancement for DR.As for the segmentation of the lungs in DR, the paper starts with classical gray-level thresholding. After the gray distribution of the lung in histogram is detected by multi-thresholding which can automatically count the class, and the images are compensated intensity according to the spatia! location of the lung and the background, a method of using largest threshold from multi-thresholding to segment the compensated images is proposed. As for the segmentation of the rib cage in DR, making full use of that the pixel in the rib cage has higher intensity in local area, it regards the smoothed image by a Gaussian filter as the threshold surface for segmentation. And it produces perfect results. Later the Gaussian filter is substituted with an equivalent mean filter, and the computation time has been reduced significantly.Image denoising and enhancement technologies for DR in this paper are concentrated on wavelet analysis . Based on discussion and explication the reconstruction from modulus maxima, phase filter and Lipschitz exponents, a method to improve the denoising method proposed by Mallat is presented, which still uses the ideal of modulus maxima reconstruction. The relating wavelet coefficients, however, are selected bv phase filter and Lipschitz exponents. The denoising results using this method show that the details in DR are kept.Unsharp masking is one of the most popular methods to medical image enhancement. Its equivalent in wavelet field is deduced in this paper. According to the principle of visually receptive psychology, the gain of enhancement should be determined by the background of the image. At the same time, the gain coefficients should be adjustedwhen taking into account the trait of DR with important details lying in medium spatial frequency. The experiment results show that wavelet adaptive enhancement with adjusted gain is effective for the enhancement of the detail information in DR.