Salient target detection algorithm based on boundary prior and iterative optimization

• Main Authors: ZHANG YEDA, ZHOU YUAN, MAO AILING, LI CHUO, YANG JING, HUO SHUWEI
• Format: Patent
• Language: Chinese; English
• Published: 17.08.2018
• Subjects: CALCULATING; COMPUTING; COUNTING; HANDLING RECORD CARRIERS; PHYSICS; PRESENTATION OF DATA; RECOGNITION OF DATA; RECORD CARRIERS

The invention discloses a salient target detection algorithm based on boundary prior and iterative optimization. The method comprises the steps of extracting feature image information, and expressingthe image information as a form of a feature matrix; establishing a region background likelihood estimation model based on boundary prior, wherein the position and the contour of salient targets can be accurately detected through the model; generating a significant graph enhancement model based on iteration optimization, namely, iteratively executing background/foreground seed selection and significant value global optimization. Compared with the prior art, the method is based on boundary prior and an iterative optimized salient target detection algorithm and combines a plurality of salient features and clues, and the salient image quality of any accuracy can be greatly improved. 本发明公开了种基于边界先验和迭代优化的显著目标检测算法，步骤、提取特征图像信息，并将图像信息表示为特征矩阵的形式；步骤二、建立种基于边界先验的区域背景似然度估计模型，通过该模型可准确检测出显著目标的位置和轮廓；步骤三、生成基于迭代优化的显著图增强模型，即迭代地执行前景/背