Segmentation and noise removal on thermographic images of hands

• Type: Article in International Scientific Journal
• Date: 2008

Authors

Vardasca, R (Author) Other; Bajwa, U (Author) Other

Journal

Title: Thermology International

Vol. 18

Pages: 89-94

ISSN: 1560-604X

Publisher: European Association of Thermology

Indexing

Scopus - 2 Citations

Other information

Authenticus ID: P-00F-V8E

Abstract (EN): Hands are physiologically those parts of the body (together with the feet) where body radiant heat loss is highest. The temperature of the hands may be very close to the environmantal temperature and therefore may be difficult to be separated from the background in the infrared images. From all parts of the body, hands are most complex in shape and therefore difficult to segment. A correct outline, however, is needed for studying certain diseases such as arthritis, neuromusculoskeletal injuries or circulatory pathology by thermal imaging. Manual segmentation is possible but time consuming and inaccurate to reproduce. The aim of this study was to investigate which of the many automatic edge detection algorithms known from literature produce the best performance in such low contrast thermal images. Additionally, the effect of pixel noise on this process is analysed by using a homomorphic filter prior to edge detection. This filter is appropriate because it allows pixel noise produced by the imaging system to be modelled as an additive term to the original image. Two analyses are performed, a visual (subjective) and a quantitative (objective) for the extracted edges. Both assessment methods conclude that the best outlining results are achieved when using a probabilistic (Canny and Shen-Castan) edge detector together with homomorphic noise filter pre-processing.

Language: English

Type (Professor's evaluation): Scientific