Automatic Thresholding for Segmentation in Chest X-Ray Images Based on Green Channel Using Mean and Standard Deviation
V.Thamilarasi1, R. Roselin2
1V. Thamilarasi, Lecturer in Computer Science, Sri Sarada College for Women (Autonomous), Salem – 16, India.
2Dr. R.Roselin, Associate Professor of Computer Science, Sri Sarada College for Women (Autonomous), Salem – 16, India.
Manuscript received on 02 June 2019 | Revised Manuscript received on 10 June 2019 | Manuscript published on 30 June 2019 | PP: 695-699 | Volume-8 Issue-8, June 2019 | Retrieval Number: H6793068819/19©BEIESP
Open Access | Ethics and Policies | Cite | Mendeley | Indexing and Abstracting
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC-BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Abstract: Threshold based segmentation plays important role in digital image processing. It is one of the best and generally used methods. Presently medical diagnosis need more support from Computer Aided Diagnosis (CAD), particularly the chest X-ray images for lungs. Segmentation of lung portion from chest X-ray plays an important role in identification of abnormality in lung region. There are lot researches going on, in the segmentation of lung region based on threshold. The universally accepted Thresholding techniques Otsu, Niblack and Sauvola methods are experimented for 247 chest X-ray images from Japan Society of Radiological Technology (JSRT) dataset[1]. The result of these methods however fails to segment lung region with good accuracy. Every image needs proper threshold for segmentation. It is difficult to fix single threshold for all images. The chest X-ray images need absolute threshold for segmentation, within a point difference which results merging of lung portion or leads improper lungs portion. This paper presents efficient Automatic threshold based on mean and standard deviation which works centered on all three channels and suggested that green channel in RGB images works well than other two channels blue and red. Based on this concept the proposed method was developed and experimented on all 247 images. This method is able to segment lung portion in all images except one which is JPCLN044. The accuracy of this method catches with the help of dice coefficient similarity method. The proposed method achieves nearly 79.71% success rate before applying filter and 79.73% after applying filter with the support of dice similarity method.
Keyword: Threshold algorithms, Mean, Standard Deviation, Green channel, Segmentation, Automatic Thresholding.
Scope of the Article: Standards for IoT Applications.