纹理和边缘特征相结合的图像修复算法

针对Criminisi算法在图像修复过程中易将纹理部分误认为边缘部分,从而造成修复顺序偏差而影响修复效果的不足,以及由于在已知区域中不存在合适匹配块而造成的修复效果失真问题,提出一种基于纹理和边缘特征相结合的图像修复算法.首先利用图像的纹理特征和边缘特征引入差别因子,改进优先权模型来增强对边缘部分的辨别能力,以克服修复过程中易造成纹理延伸的不足;通过改进原算法的搜索方式,在提高算法效率的同时去除修复过程中的瑕疵点;然后对修复过程中所产生的不良匹配区域进行定位,采用基于图像分解的图像修复算法进行修复.实验结果证明,改进算法较好地克服了Criminisi算法存在的缺点,能获得较好的视觉效果.

Identification of differentially expressed genes in human uterine leiomyomas using differential display

In searching of differentially expressed genes in human uterine leiomyomas, differential display was used with twelve pairs of primers to compare human uterine leiomyomas with matched myometrium. False positives were eliminated by reverse Northern analysis. Positives were confirmed by Northern blot analysis. RESULTS: Four of 69 cDNA fragments (3 up-regulated named L1, L2 and L3 and 1 down-regulated named M1 in leiomyoma) were confirmed by Northern analysis. Sequence comparison and Northern analysis proved that L1 is exactly the human ribosomal protein S19. It was present ubiquitously in 13 tissues tested but in various levels and even in different size. L1 was highly expressed in parotidean cystadenocarcinoma, pancreatic cancer and breast cancer examined. No mutations have been found in human uterine leiomyomas (n=6). CONCLUSIONS: hRPS19 overexpression might be a universal signal in rapid cell growth tissues.

利用纹理和边缘特征的Criminisi改进算法

针对Criminisi算法在图像修复过程中易将纹理部分误认为边缘部分,造成修复顺序偏差而影响修复效果的不足,充分利用图像的纹理结构特征和边缘结构特征,引入差别因子,改进优先权模型,以增强对结构部分的辨别能力,并通过采用新型的搜索方式来克服在修复过程中易产生瑕疵点的不足来完善修复效果。实验结果表明,改进算法较好地克服了原算法所存在的纹理延伸等不足,保持了修复内容的线性结构,其峰值信噪比相比于原算法提高了2～3 dB,具有更好的视觉效果。