This technique was developed for a company looking to detect unhealthy and aged skin using a number of computer vision algorithms and for the recommendeation of creams. The algorithms have been optimised for mobile devices.

Introduction

Wrinkle detection algoithm using a number of filters including gabor, frangi and a novel thresholding method. This is followed by a novel line tracking algorithm to detect the biggest wrinkles. The algorithm was tested on the AGES dataset with an accuracy of 79.41%. It is mentioned in more detail in the following publication:

Joint Roughness and Wrinkle Detection Using Gabor Filtering and Dynamic Line Tracking (https://www.cscjournals.org/library/manuscriptinfo.php?mc=IJCSS-1507)

Description

A number of filters are applied in order of gray-scaling, gabor filtering, and frangi filering to find lines in the provided image, in this case wrinkles. The images below show the image before and after using the edge detection filters.

After this process a novel vertical thresholding and blob removal algorithms are applied to the technique to remove any horizontal wrinkles and any areas below a certain size.

This is followed by a novel line tracking algorithm to remove wrinkles that are below a certain length and to keep the most dominant. The images below show the first filtered image and after the thresholding, dot removal, and line tracking is applied.