Abstract: To solve the problem that the salient detection method has low contrast, the object area is not obvious, the detection area is not accurate, and the background suppression effect is insufficient. The power object transformation and IGLC algorithm are proposed. First, the power law transformation function is used to optimize the IG algorithm to completely suppress the background area of the saliency map. The binarized saliency map is segmented in the original image to obtain the segmentation map of the object of interest. Secondly, the LC algorithm optimizes the segmentation map of the object of interest to obtain the saliency map with good detail. Finally, the adaptive fireworks algorithm is used to enhance the contrast of the salient object region, generating the final saliency map. The object test of the standard test dataset MSRA10K and PASCAL-S datasets is performed, and subjective and objective comparison analysis is carried out with the six popular object detection methods. The analysis results are superior to the comparison methods. The salient map obtained by the algorithm has the advantages of contrast and detail enhancement, and has better background suppression effect.

Key words: salient object detection, power law transform, IG algorithm, Luminance Contrast（LC）algorithm, adaptive image enhancement

摘要： 针对显著性检测方法生成显著图存在对比度低、目标区域细节不明显、检测区域不准、背景抑制效果不足的问题，提出幂律变换和IGLC算法的显著性目标检测方法。利用幂律变换函数优化IG算法，彻底抑制显著图的背景区域。经二值化处理的显著图在原图像分割，得到感兴趣目标分割图；LC算法优化感兴趣目标分割图，得到细节佳的显著图；利用自适应烟花算法增强显著目标区域的对比度，生成最终的显著图。对标准测试数据集MSRA10K和PASCAL-S数据集中的图像进行显著性目标检测实验，且与目前较流行的6种显著性目标检测方法进行主观和客观的对比分析，分析结果均优于对比方法。该算法得到的显著图既具有对比度和细节增强的效果，又具有背景抑制效果更好的优点。

关键词: 显著性目标检测, 幂律变换, IG算法, LC算法, 自适应图像增强