THE VISUAL REQUIREMENTS OF EVERYDAY TASKS

传播学理论将应用于视觉的研究 系统： o制定一个视觉敏感度图表，“绕过”光学， the eye. 噪声中的扰动对光学缺陷非常不敏感。 我们将测试和进一步开发一个字母在噪音测试图表 它利用这一事实来评估视觉表现， 存在光学受损。 理论和试点数据表明， 性能几乎完全不受光学缺陷的影响， 这意味着噪声字母低于正常值的患者 敏感性必须有神经缺陷，无论是视网膜的还是中枢的。 测试图具有评估视力的潜力，即使在 存在致密性白内障，并可能具有诊断价值 区分不同种类的自然缺陷。 o测量视觉注意的信息容量。 我们假设视觉注意的信息容量 可以用信息容量的多少来表征 在显示器所需的最佳性能的一个细心的 观察员的任务。 我们预测这种信息容量 对于所有视觉任务， 需要注意。 12个实验将检验威尔假设 通过测量各种各样的所需信息容量， 视觉任务。 o从运动中提取结构。 我们将研究人类的运动和深度知觉， 机器视觉系统，当图像受到低 分辨率低、对比度低、视野小或时间模糊。 我们将使用现有的运动算法来提取三维 结构从二维速度流场。 人类 在同一任务的表现将被比较， 算法 结果将测试现有的运动恢复结构 作为人类运动和深度感知模型的算法 视觉系统

Communication theory will be applied to the study of the visual system to: o Develop a visual sensitivity chart which "bypasses" the optics of the eye. Thresholds in noise are remarkably insensitive to optical deficits. We will test and further develop a letters-in-noise test chart which exploits this fact to assess visual performance in the presence of impaired optics. Theory and pilot data show that performance is almost totally unaffected by optical deficits, implying that patients with below normal letter-in-noise sensitivity must have neural deficits, either retinal or central. The test chart has potential for assessing vision even in the presence of dense cataracts, and may have diagnostic value in distinguishing different kinds of natural deficit. o Measure the information capacity of visual attention. We hypothesize that the information capacity of visual attention can be characterized by how much information capacity is required at the display for optimal performance of an attentive task by the observer. We predict that this information capacity will be constant, on the order of 100 bits, for all visual tasks requiring attention. Twelve experiments will test will hypothesis by measuring the required information capacity for a wide variety of visual tasks. o Extract structure from motion. We will investigate motion and depth preception of human and machine vision systems when images are restricted by low resolution, low contrast, small visual field, or temporal blurring. We will use existing motion algorithms to extract 3-dimensional structure from a two-dimensional velocity flow field. Human performance at the same task will be compared to that of the algorithm. The results will test existing structure-from-motion algorithms as models of motion and depth perception in the human visual system.