Lazy coloring of images with closed regions

Abstract

Adding colors to images is a common operation in image manipulation. Many works havebeen proposed to facilitate this colorization process targeting at different types of input images. However, these methods require users to indicate regions by precise clicks or strokes, which is a tedious process and difficult to operate on touch devices due to the well-known fat finger problem. In this thesis, we propose a new tool which is able to determine meaningful regions when the user draws a rough stroke to colorize an image. The input image is composed of well-defined closed regions, which can be a template image, or extracted from a cartoon or line drawing image by segmentation. The correlations between pairs of regions are pre-analyzed before editing. When a stroke is drawn, the possible intention of the user is estimated by grouping regions that share common properties. By matching the location and shape information between candidate groups and the stroke, the candidate groups are ordered to capture their possibilities to be the intended colorization regions. The group with the best score is returned as the default colorization result, and an ordered list of candidate groups is displayed for possible selection of alternative colorization results. Our tool allows regions that share common properties to be colorized using a single stroke, drawn only roughly, making our tool suitable for touch-based interfaces. We conducted a user study to demonstrate the convenience of our tool.

Date of Award	2012
Original language	English
Awarding Institution	The Hong Kong University of Science and Technology