Study on feature extraction and filling-in process in visual system

(1)我们尝试将可以解释某些亮度错觉的模型应用于实像，以检验视觉系统中产生视错觉的原因。结果表明，产生亮度错觉的机制与双眼视网膜图像的聚焦和融合有关。(2)视网膜锥在光适应中起重要作用。为了理解光适应，有必要考虑锥体的特性。锥体的光强响应曲线为s形曲线。我们利用这条曲线来增强数字图像的对比度。该方法可以改善以往增强方法存在的一些问题。(3)人类视觉系统存在整体（平均）适应和局部适应。此外，一项生理实验表明，人类视觉系统可能存在中程适应。本研究通过比较两个斑点的主观亮度来检验中距离适应的存在及其特征。结果表明，当亮度阶跃变化存在于中央凹处时，中距离适应发生，而存在于中央凹外时则不发生。(4)为了了解主观圆是否比背景亮，我们测量了面向线端部分和面向条纹平行侧部分的感知亮度。随着线间距的减小，线端（平行侧）的感知亮度变亮（变暗）。(5)对高动态范围图像（60dB以上）采用中程自适应和锥的s形曲线。提高了明区和暗区对比度。

(1) We tried to apply the model which can explain some brightness illusions to the real image in order to examine why visual illusion occurs in the visual system. As a result, it is suggested that the mechanism which produces the brightness illusion is concerned in focusing and fusion of the binocular retinal images.(2) Retinal cones play an important role in light adaptation. To understand the light adaptation, it is necessary to consider characteristics of the cone. The light intensity-response curve for the cone is a S-shape curve. We employ this curve to enhance a contrast of digital images. This method can improve some problems of previous enhance methods.(3) Overall (average) adaptation and local adaptation exist in human visual system. In addition, a physiological experiment suggests that the midrange adaptation may occur in human visual system. The present study was carried out to examine the existence and the characteristics of the midrange adaptation by comparing subjective brightness of two spots. The results indicate that the midrange adaptation occurs when a step change in luminance exists on the fovea, but not when it exists out of the fovea.(4) In order to understand that the subjective circle is brighter than the background, we measured the perceived brightness of the part which faces line end and part which faces the parallel side of the stripe. With decreasing line interval, the perceived brightness of the line end (parallel side) is brighter (darker).(5) We apply the midrange adaptation and S-shape curve of the cone to high dynamic range images (over 60dB). Both contrast of bright region and dark region was improved.