Ukrainian Company Creates New Algorithm for Digital Image Processing

The Ukrainian software development startup company, Image Upscaler, has created a new image processing algorithm that can allow anyone to remove the backgrounds from their digital images over the internet. Image Upscaler is releasing this new digital software as an online service to all users worldwide. 

“It works as an online service,” said Evgen, a representative of Image Upscaler. “Users can simply upload their images on our webpage and then download the new images without backgrounds in them. The whole conversion process works in a matter of seconds. Our software replaces the old backgrounds with an all-white color, which creates a ‘green screen-like effect’ that allows users to replace the white background with whatever images they want without disturbing the foreground.”

Image Upscaler developed the algorithm based on Deep Learning, which is a unique class of automated machine learning algorithms with several different layers. As the algorithms absorb new visual and textual data automatically, its artificial intelligence acquires new knowledge based on the actions of humans. 

The specific algorithm associated with the background removal software has two separate parts. The first CNN detects the main objects within the image (the background and the foreground) and creates a rough mask of them. 

The second CNN receives input from the mask and then creates a new mask with a white background instead of the original background. It does this by subtracting the rough background from the input received and then uses very accurate edge cutting to establish the foreground image separately.

“Many online entrepreneurs need to have solid white backgrounds in their product images when selling things like clothes and shoes,” said Evgen. “Our online service provides an easy way to produce those white backgrounds without having to buy expensive photo editing tools which require advanced skills. Users of our service do not need to have any photo editing skills whatsoever. It is a completely automated process that does all the work for them.”