Noise Removal in Dermoscopic Images Using a Novel Software.

Sir,

Dermoscopy enables magnified in vivo visualization of subcuticular features invisible to the naked eye.[1] It may be performed manually or with digital systems such as video camera to archive the images over a period of time.[2] Presence of the hairs, ruler markings, and air bubbles may hamper the dermoscopic image's evaluation.[23] Thin hairs generally do not interfere in the diagnosis but thick hairs can be a distraction. In technical terms, these are called “noise,” and removing noise will not affect the diagnosis and may increase the diagnostic accuracy.

Assistance of certain software can aid us to remove the noise. This involves several steps. We have employed such a technique in one of our dermoscopic images of psoriasis [Figure 1a]. We used polarized dermoscopic images using Hein Delta 20 plus with Sony Cyber-shot DSC-W830 camera. Initially, original image was converted to an 8-bit color image and edge detection method was implemented to detect the artifacts; then, the morphological feature detection method was used to segment the areas detected by edge detection method [Figure 1b]. Finally, we designed our own method to find the noise region and to mask or replace that region [Figure 1c]. The following code replaced the noise region by the neighboring pixels. This enabled us to obtain a final image that is devoid of the noise and without any disturbance in the required dermoscopic features.

Figure 1

(a) Original dermoscopic image, (b) artifacts segmented image, (c) output image

The concept of noise removal is old but its utility has been explored for all dermoscopic images. Software removes thick hairs without affecting the diseased region [Figure 2]. Image enhancement algorithms implemented in the software improves the output image quality, which helps in better illustration and archival. It utilizes a user friendly interface (GUI) and faster processing using an endogenously designed software which is available on the internet. Time taken for the software for image processing ranges from 30 seconds to 3 minutes depending on the image size and presence of noise. There is no need for any sophisticated computer skills. The proposed tool will not remove hair at ill-focused region and very thick hairs; however, we are working towards incorporating these features too.

Figure 2

(a) Original dermoscopic image, (b) output image

We propose this software utility for the dermatologists to increase the diagnostic accuracy of the dermoscopy. This software is available in the university's research centre website (http://yenepoyaresearch.edu.in/Patents_softwares.php).

Financial support and sponsorship

Nil.

Conflicts of interest

There are no conflicts of interest.