Most of the lung lesions may not be detected due to the fact that they may be camouflaged by underlying anatomical structures, or the low quality of the images, or the subjective and variable decision criteria used by the radiologist. Therefore the most important and difficult task, the radiologist has to carry out is the detection and diagnosis of cancerous lung nodules from chest radiographs. These are problems that cannot be corrected with current methods of training and high levels of clinical skill and experience. The present research work describes the computerized technique to identify the lung nodules by extracting various discriminating geometrical and textural features like area, perimeter, irregularity index, standard deviation, skewness, third moment, entropy etc. using image processing and analyzing algorithms. Then these features are applied as an input to the feed forward neural network for the classification of lung cancer. Thus the developed algorithms aid the physician to detect the cancer in a short time with more accuracy.