INTERACTIVE GRAPHIC APPLICATION FOR IMAGE COLORATION

Abstract

Nowadays we have thousands of products related to image adjusting adding filters, blurring, removing noise, etc. Also, we have applications related to image coloration, however, their functionality is quite narrow. With the help of the interactive graphic application for image coloration, it would be easier to adjust, particularly recolor, fully recolor, and even color black and white images from scratch. The application considers users’ choices by interactive instruments: texting the preferred color by some phrase “once in a blue moon”, choosing colors with a color picker, choosing a palette, choosing a palette with random colors, and adjusting a palette. All these instruments are applied to both color and black & white images. An example of using this application is in urban planning or design, where it is quite taught work for architects to make up with the idea of a new color palette for the particular project. For such cases, we eve provided functionality for fully coloring some raw project in form of black & white image. The paper has the description and the results of the creation of a graphic application, the basis of which are: windows graphical user interface, mathematical clustering method K-means, and a convolutional neural network. Using of clustering method K-means deserves a separate section, as there was used not the primary method, but modified – the applying of k-means to colors of an image, each iteration of the algorithm touches every pixel in the image, it is expensive for volumetric images, thus after analyzing different approaches, I decided to choose the so-called weighted K-means approach, in which each point C represents a certain number of pixels, the array of C point is called bins. In addition to the clustering method, the system uses the recurrent neural network to process text values or generate colors. As the result, user can connect to the system through a simple windows graphical interface that allows him to upload the image for further processing. The practical value of this work is the developed high-quality application which is ready for use and implementation in the relevant fields.