| Summary: | 碩士 === 國立清華大學 === 電機工程學系 === 97 === For these years, computer vision has made great progress in pattern recognition. Many recognition models have been proposed and some of them achieved fairly good recognition rate. The ultimate goal of these methods is mainly to realize the recognition capability of human vision. Usually, for an image containing a simple object, the human vision will associate only one interpretation of the object in response to such image. However, there are cases where one viewer or different viewers may not always respond with the same answer when viewing the same simple figure. We call this multi-interpretations. About this phenomenon, we propose an explanation that it is because the viewer(s) might have different sensing criterion when viewing an image. After the differently sensed result is received and processed by the brain, different response pattern might be produced, thus different interpretations to the same image occur.
To support our idea of explaining the aforementioned phenomena, we develop a pixel resonance concept based on random sampling and the photon cluster resonance concept of quantum mechanics to model different sensing criterion effect. A Monte Carlo method is applied in our pixel resonance simulation to illustrate the proposed multi-interpretations concept. If a given image is randomly sampled by different sampling energy, different resonance results might be obtained, thus multi-interpretations.
We then extend the pixel resonance and multi-interpretations concept to color image and color blindness tests. Using different resonance conditions, the different resonance images are obtained. In addition, we further explain the concept about multi-interpretations by simulating color blind vision of color blindness tests.
|
|---|
