Sputum is a liquid substance or mucus produced in the lungs and in the airways leading to the lungs. Sputum can be categorized in a type of bloody that often found in tuberculosis (hemoptysis), rusty colored where it can be found in pneumonia, purulent, foamy white and frothy pink. A patient suspected having respiratory infection could be examined by sputum diagnosis. Pulmonary infection is the most cases for the respiratory infection and it can cause of morbidity and mortality. Sputum sample that contain of pus cells, squamous cells, gram-positive and gram-negative organisms are usually examines by microbiology technologist using visualize inspection through microscope image. It is less practical to examine the huge sample of sputum by manually visualizing image under the microscope. Hence, it is crucial to introduce an automatic system using image processing technique where it can overcome the inefficient practice of examine the sputum sample. In this paper we proposed an image processing technique to detect the types of cells in sputum smear. The image processing methods used to analyze the cells are color, morphology, image arithmetic operation, K-Mean clustering and thresholding technique. The sputum image has been tested for the cell detection where we obtained more than 95% detection of squamous cell.