<p>Attention retargeting is a technique employed to manipulate images and guide viewers’ attention toward specific regions of interest. However, traditional methods such as recoloring the target region often produce unnatural and esthetically unpleasing results due to limited control over semantic consistency and visual appeal. In this paper, we propose a novel framework for palette-based image editing and attention retargeting. Our approach uses color palettes to guide attention while maintaining visually appealing compositions. We generate a collection of semantic-aware palette dictionaries by extracting color palettes from reference images. These dictionaries are combined with a multi-constraints optimization process to determine an optimal replacement color palette. Using this palette, we recolor the target region, enhancing visual attention while preserving quality and naturalness. Experimental results demonstrate the effectiveness of our method in achieving more natural and esthetically pleasing outcomes.</p>