Enhancing saliency of a target object through color modification of every object using genetic algorithm

Abstract

Ability to draw an observer’s gaze toward an object in an image depends on the visual saliency of the object. Contrast and/or uniqueness of low-level features of an object (e.g., intensity, color, etc.) drives visual saliency in the absence of external influence. Hence, an increased difference between a user-defined target object (whose saliency is being enhanced) and its surrounding objects in terms of these low-level features enhances the visual saliency of the target object. In this paper, enhancement of saliency of the target object has been posed as a maximization problem. Feature values of every segment in an image have been modified within a limit to maximize the feature difference between the target segment and all other segments. The imposed limit restricts the amount of change in any segment in order to preserve the naturalness of the image. In the framework of the proposed maximization problem, a genetic algorithm-based approach has been devised to find suitable features values for every segment. Effectiveness of the proposed method has been demonstrated through suitable objective measures. © Springer Nature Singapore Pte Ltd. 2019