Logo image
Discussion for the Stochastic Resonance Effect on Human Visual Perception Created with Strength Variation of R, G, and B Part in Color Images
Thesis

Discussion for the Stochastic Resonance Effect on Human Visual Perception Created with Strength Variation of R, G, and B Part in Color Images

Chai, Chih-Yao
Masters, 國立清華大學, 電機工程學系
2010

Abstract

隨機共振 人類視覺 彩色影像 影像品質評估模型 視覺測試 Stochastic Resonance Human Visual Perception Color Image Image Quality Estimation Model Test for Visual Perception
Noise in nature is often seen to destroy the original signal without any benefits. Because the concept of the Stochastic Resonance has been proposed, the fact that noise can enhance signal recognition is widely concerned and researches in many fields of signal processing are started with this way for the extension of application. Our study is mainly focused on discussion for the Stochastic Resonance effect on human visual perception created with strength variation of R, G, and B part in color images. This thesis is the extension of the previous researches of Stochastic Resonance related to gray-level image and human visual perception. We will begin our research with the explanation of Stochastic Resonance and the quantization for human visual perception then introduce the image quality estimation models used in this thesis. After introducing those models, the experiment procedure will be presented. There are two experiments in this thesis. In the first experiment (experiment A), original natural color images and luminance-shifted are corrupted with various noises and thresholds. The comparison of their results will then be done to verify the fact that weak image is easier to cause the SR effect for human visual perception. For the second experiment (experiment B), we transform the gray-level test image used in previous researches to RGB color space with different color background then make the test images used in this thesis with various noise-threshold. Those test images are used to create the test videos. Ten human subjects are asked to evaluate the test videos and their perception results will be averaged. The image quality results for those test images will be evaluated with image quality estimation models. We present the results with tables and do the analysis and discussion in the text of this thesis. For the results of subjects, we summarize their properties and do the comparison. For the results of image quality estimation models, we perform the inference individually with their properties or mathematical functions. In this experiment, in addition to the image quality estimation models used in previous studies, we will use three other models to make the application wider.

Metrics

1 Record Views

Details

Logo image