Skip Ribbon Commands
Skip to main content

Getting Rid of Dark Spots


Out, dark spot! A new computer program by the School of Electrical and Electronic Engineering (EEE) at Nanyang Technological University (NTU) could help to erase literal dark patches on people’s health. Professor Lin Zhiping and Professor Ser Wee at Centre for Bio Devices and Signal Analysis (VALENS) have developed an image processing technique to help doctors better assess patients with melasma, a skin disease that causes dark patches on the face, and determine their treatment’s effectiveness. 

Melasma affects millions of people, but dermatologists currently manually examine the patients, which means the diagnosis and treatment depends on the clinicians’ experience and is subjective. The EEE image processing technique objectively analyses images of patients’ faces, and produces a Melasma Area And Severity Index based on the percentage and darkness of the melasma regions. 

The researchers combined global and local thresholding, two techniques used to segment images, for their hybrid thresholding method. Global thresholding looks at the image as a whole and selects a brightness threshold to divide all pixels into two categories – brighter or darker than the threshold. This is useful for picking out areas that are distinctly darker than others, such as very dark melasma regions. 

The method, however, can mistakenly disregard small melasma spots as image noises. It can also be misled if the face is unevenly lit, since shadows cast by, say, the nose can artificially darken some areas. Local thresholding compensates for such errors. Each pixel is compared to its neighbours in a preset area before a threshold is set to categorise it. Dividing the image into multiple overlapping areas means each one is more evenly lit even with shadows across the face. 

Professor Lin Zhiping from EEE said local thresholding also has weaknesses. If a preset area has little internal variation – for example, if it is entirely within an even melasma region – the technique can misclassify it since there are few or no different pixels for comparison. He said: “We combined both methods for a better result, and our work can help patients get better diagnoses and treatment.”


ByProf Lin Zhiping,  Prof Ser WeeSchool of EEE


Click here to find out more.

Not sure which programme to go for? Use our programme finder
Loading header/footer ...