工业CT重建图像中边缘梯度效应引起的高频伪影

为研究伪影对大型工件内部密度锐变区域高对比度微小缺陷检测的影响,从一个简单的轴对称断层模型出发,计算了边缘梯度效应引起的通过物体内部不同材料交界面处的射束的投影偏差。并研究采样相位的影响,探讨了由此偏差引起的重建图像中的高频伪影的表现形式。计算结果表明,高频伪影具有相当的强度,且具有与缺陷相似的表现形式,从而影响工业CT的检测能力。结合点扩散过程,高频伪影强度的计算为评估工业CT边缘处的缺陷检测能力提供了理论依据。     

基于变换域注意力机制的压缩伪影去除的方法

图像有损压缩过程往往会导致图像质量退化,使图像出现压缩伪影。针对现有基于深度学习的方法缺乏对联合图像专家组(Joint Photographic Experts Group,JPEG)压缩算法先验信息的利用,提出一种基于变换域注意力机制的去伪影方法。该方法利用卷积神经网络在像素域和离散余弦变换(discrete cosine transform,DCT)域分别提取特征,再将双域学习的特征信息进行融合。利用量化表设计了DCT注意力机制,该模块根据DCT系数的损失程度给予各频率系数不同的权值,使网络自适应补偿量化引起的误差。于此基础上,在像素域引入通道注意力机制,从而更好地利用量化表的先验信息。在主要数据集上,提出的去伪影方法以固定的模型参数对多种质量因子的压缩图像进行伪影去除实验。实验结果表明,所提出的方法在各评价指标和主观视觉上取得较好的效果。     

基于投影的CT图像金属伪影非线性权重校正

为抑制计算机层析造影(CT)系统重建图像中出现的金属伪影,提高图像的质量,采用一种基于金属投影非线性权重衰减的方法进行校正。对原始的投影直接重建,在得到的图像中利用阈值分离出金属区域,并重新对该金属区域进行投影,便得到原始的投影中金属投影的区间。只对该金属投影进行合适的非线性衰减。再用传统的滤波反投影方法重建图像。数值模拟表明,该方法不但能基本消除由射线硬化和射线剧烈衰减而引起的金属伪影,而且能保留金属信息,同时能增强金属区域不同部分之间的对比度,使得重建图像的质量得到很大的提高。该算法计算复杂度较小,实现简单,具有较高的实用价值。     

红土洞遗址石制品的初步研究

本文报道贵州省红土洞史前遗址出土的石制品,共有56件,包括石锤(7件)、石核(7件)、石片(13件)、石器(29件)等类型。石器类型简单,只有砍砸器、刮削器、盘状器3种。石器尺寸相对穿洞出土的石器大,石器修理较细致,器形精美。     

关于技术进化的个案研究

人工制品是理解技术的重要维度之一。人工制品的形成和发展是一个进化的过程，其“起源’’于已有的技术，发明的动力源于技术体系与人的耦合所决定的技术趋势；人工制品的整个生命周期是一个贯穿着技术趋势的不断具体化的过程，进化的实质在于人工制品在具体化的过程中获得作为存在的意义和厚度。     

超分辨率处理中振铃现象的分析与抑制

为了抑制在实施FFT插值法重构的超分辨率图像中可能出现的振铃现象 ,深入分析了出现振铃现象的原因及产生条件 ,在此基础上设计了抑制振铃的频域补偿滤波器及其自适应算法。大量实验和比较结果表明 ,处理后的图像不但分辨率有所提高 ,而且未出现振铃现象 ,质量更好 ,从而证明了所设计的抑制振铃频域补偿滤波器及其自适应算法的有效性。     

The specific contribution of object''''s origin on artifacts categorization

Gelman and Bloom found that adults and children's object naming was sensitive to how an object was created (man-made or not), but they did not reveal on which specific level of conceptual system this effect was. Using a free-naming task and a force-choice task, two experiments were conducted to test a hypothesis that this effect was specifically on domain level ("artifact/non-artifact" distinction). In Experiment 1, participants were asked to name shortly-depicted objects, rate their confidence, and report their reasons for each naming response. Results showed that most of the naming responses in "man-made" condition were in artifact domain, and most in "natural" condition were in non-artifact domain, although in both conditions names were very divergent on basic level. In Experiment 2, another group of participants were asked to choose one from two names (one in artifact domain and the other in non-artifact domain) to match the same shortly-depicted objects presented in the first experiment. Re-sults of Experiment 1 on domain level were replicated in Experiment 2. These convergent findings supported the hypothesis that the effect of object's origin is specifically on domain level of conceptual system of objects. Reasons explicitly reported for naming responses in Experiment 1 suggested that participants might automatically infer objects' functions in "man-made" condition but not in "natural" condition. Here the function-based hypothesis of artifacts classification is discussed.     

基于数学形态学的去方块效应法

为了减少多分辨奇异值分解(MSVD)图像融合后的图像出现明显可见的方块效应,提出一种基于数学形态学的边缘检测去方块效应的方法.首先通过形态学梯度算子求出两幅原图像的边缘,将边缘融合后的图像作为有方块效应图像的边缘,然后根据形态学梯度逆变换得到结果图像.最后,采用客观性能指标对结果图像进行评价.实验结果表明本文中的方法不仅简单易行,而且图像可以达到较均匀的成像效果;方块效应去除后,图像表现出良好的视觉效果,清晰度和空间频率都有很大提高,且具有较好的保持边缘的性能.     

基于二维尺度维纳后滤波器的SFCW-MIMO雷达成像伪影抑制方法

为了抑制传统后向投影成像方法产生的伪影,提出了一种基于二维尺度维纳后滤波的伪影抑制方法。首先,针对步进频连续波MIMO雷达,将频域BP成像方法建模为一个二维无失真波束形成器;然后,利用带尺度因子的二维维纳后滤波器使波束形成结果的均方误差最小;最后,分析了尺度因子对单目标和多目标成像性能的影响。现有基于相干因子(CF)的伪影抑制方法仅是所提方法具有不同尺度因子时的特例。通过选择适当的尺度因子,该方法可以克服CF方法的一些局限性。仿真结果表明,所提方法在距离向和方位向都能有效地抑制成像过程产生的伪影。     

Naturally-formed objects categorized as artifacts: Effect of objects’ functional depictions

Past research has identified a variety of factors that influence cognitive processing of artifact concepts, but few experiments tested possible interactive patterns among these factors. Using a natural language depiction based free naming task, we explored how an object’s function and its history (information about how an object was formed by natural force) interact to affect categorization judgment in a 2×2 between- subjects experiment (N = 169). The study had two key findings: (i) Without presenting an object’s function, presenting the history of its natural formation increased subjects’ tendency to name the object under the “natural” domain and decreased their tendency to name it under the “artifact” domain; (ii) once we presented the depiction of an object’s function, the effect of history presentation on categorization disappeared: Subjects in both conditions strongly preferred naming objects as artifacts. In summary, presenting depiction of an object’s function inhibited the effect of history on object categorization. Current findings suggest that an object’s function plays a fundamental role in the cognitive processing of artifact concepts.