Bilateral functions of direction and pattern by auto-correlation model-categorizing texture and office images
Tatsuya Shibata, Takumi Kato · 2002
It is known that visual functions such as color, direction, and movement are specialized in the brain. However, it is not clear how the functions are detected. We introduce an auto-correlation model to detect direction parameters. The paper examines how important pattern is as a visual function and how it is detected. We apply an auto-correlation model because direction is based on continuity, while pattern is based on discontinuity. We therefore hypothesize that pattern is an opposite function against direction. Pattern (in the article), means a repetition of different areas, which has a property of discontinuity. We conclude that the auto-correlation model proposed detects not only edge and line directions but also patterns, repetitions of different areas in images, and is useful for discriminating texture images by material and retrieve images by words, which are supported by experiments.