視覺目標感知程序中亮度、對比與梯度結構之基本效應觀察

Abstract

本論文旨在觀察利用亮度、對比與梯度之間的交互作用，於灰階影像中擷取視覺目標的區塊結果，共分成兩個觀察項目：亮度與梯度結構的交互作用、對比與梯度結構的交互作用。

本論文將亮度與對比分成強、中、弱三個部分，分別與梯度結構的同向性分佈區域、逆向性分佈區域及散亂性分佈區域交互作用，產生一系列的目標擷取圖，以觀察其效應。

利用一般化的影像特性，透過不斷的實驗與觀察，探討亮度、對比與梯度結構對目標擷取的結果與主觀認知的差距。

In this thesis, experiments on using the interactions among intensity, contrast and gradient organization(GORG hereafter) for capturing visual targets in a graylevel still image will be reported.

The GORG, gradient organization, at a specific locality is an index reflecting whether gradients in the neighborhood are mostly in alignment with, opposite to, or in randomness with respect to the central gradient under consideration. For any graylevel image, an associated GORG map can be derived, as well as a contrast map too.

Two observations were conducted during the experiments. The first focused on the interactive behavior of intensity and GORG attributes in the product of the image and GORG map; the second on that of the contrast map and GORG map.

Specifically, the another is interested in (1) The results from the interaction between pixels of high, medium, and low intensity(contrast) along with those of alignment, opposed, and random GORG(consequently a outcomes from each of the two observations for every image input). (2) Getting a preliminary picture of how much the deviation from the subjective visual target perception may be using the three cues proposed in this thesis for capturing visual target by computation.