Image Segmentation for Lung Cancer Detection Using Gabor Filters

The process of identifying lung cancer from low quality X-Ray images has been discussed widely in medical sector. The problem of false positive results in identifying cancer introduce requirement of more sophisticated method. We propose a robust feature estimation and ranking method for the segmentation of lung images using which LCD can be performed. In the proposed approach the input image is filtered with gabor filter then, the lung region is identified and extracted using template matching techniques. The extracted lung region is sub sampled into number of small images called integral images. For each sub sampled integral image, we identify set of interest points where there is more valued pixel present in the integral image. The feature descriptor for each interest point is computed to represent the features around a point and integral image. The white matter deapthness is estimated using the gray values of the feature descriptor and ranked according to them between various integral images. Finally a top ranked region or integral image is selected and marked as identified location of LC. The proposed method reduces the time complexity and false positive results and produces efficient results.