CAJ | 학술논문

针对红外成像制导复杂背景下低对比度红外图像的分割问题，提出了一种新的基于Kapur最大熵阈值判别式的二维斜分快速递推算法，并采用逐步逼近的粗细搜索策略，减少阈值搜索区域，在可能的阈值范围内逐点寻找最佳阈值。通过对算法的复杂度进行分析，并对实际获取的红外图像进行分割实验表明，Kapur最大熵阈值判别式更加适合于低对比度红外图像分割，提出的二维斜分快速算法所需的运行时间和存储单元均少于现有的二维直分或斜分最大熵分割快速递推算法，运行时间约为原始算法的14%，分割结果的噪点更少，边界更加细致完整，适用性更强，满足红外成像制导系统工程实用化要求。
침대홍외성상제도복잡배경하저대비도홍외도상적분할문제，제출료일충신적기우Kapur최대적역치판별식적이유사분쾌속체추산법，병채용축보핍근적조세수색책략，감소역치수색구역，재가능적역치범위내축점심조최가역치。통과대산법적복잡도진행분석，병대실제획취적홍외도상진행분할실험표명，Kapur최대적역치판별식경가괄합우저대비도홍외도상분할，제출적이유사분쾌속산법소수적운행시간화존저단원균소우현유적이유직분혹사분최대적분할쾌속체추산법，운행시간약위원시산법적14%，분할결과적조점경소，변계경가세치완정，괄용성경강，만족홍외성상제도계통공정실용화요구。
To deal with the low contrast infrared images in complex background for infrared imaging guidance, a new fast recursive method based on Kapur's maximum entropy threshold discriminant was presented, coarse-fine searching strategy with successive approximation was adopted to reduce the threshold searching area, and the best threshold was searched by pixel in the possible area. The analysis of the methods' complexity and the segmentation experiments of the real infrared images show that, Kapur's maximum entropy threshold discriminant is more suitable for low contrast infrared images'segmentation, the running time and memory cell needed by the presented method are all less than existing fast maximum entropy threshod recursive methods based on two-dimensional histogram vertical or oblique segmentation, the running time is about 14% of the original method. In the result image, the noise is less, the boundary is more elaborate and complete, the applicability is stronger. The presented method meets the engineering practical requirement of the infrared imaging guidance system.