针对图像合成结果易受颜色取样和权值选取影响等问题, 本文借鉴梯度域和线性alpha合成算法的优点, 提出了一种基于采样抠图和自适应颜色权值的图像合成算法。算法首先通过增加数据项约束和邻域限制对原始抠图方法得到的alpha matte值进行局部迭代修正, 根据得到的alpha matte值对合成区域进行划分, 不同的合成区域采用不同的合成策略, 边缘区域采用梯度域合成方法对边界进行无缝对接和颜色保真, 同时利用均值插值重建方程求解, 降低求解复杂度;内部区域通过亮度对比系数计算自适应颜色权值, 并用其进行内部区域颜色再平衡, 保证源图像和目标图像的完美融合。仿真实验表明, 该算法与其他合成算法相比, 在无缝边界的同时, 颜色保证处理更为有效。

As for the problem that image synthesis result is subject to color and weight selection, the paper proposes an image synthesis algorithm based on sampling cutout and adaptive color weight by combining the advantages of gradient domain and linear alpha synthesis algorithm. This algorithm increases data item constraints and neighborhood restrictions to realize the local iteration correction of alpha matte value obtained through the original cutout method. The synthesis area is divided according to the obtained alpha matte value. Different synthesis strategies are adopted for different synthesis areas. Gradient domain synthesis method is used for seamless docking and color fidelity of borders. Besides, an equation is re-created with mean interpolation and then the solution is obtained. Thereby, the solving process is simplified. For the internal area, the adaptive color weight is calculated with the use of brightness contrast factor. Besides, the colors of the internal area are further balanced so as to ensure the perfect integration of source image and target image. The simulation experiment shows that the algorithm achieves more effective color maintenance while creating the seamless border than other synthesis algorithms.