Three-Dimensional Film Image Classification Using an Optimal Width of Histogram

Abstract

Three-dimensional film images which are recently developed are seen as three-dimensional using the angle, amount, and viewing position of incident light rays. However, if the pixel contrast of the image is low or the patterns are cloudy, it does not look three-dimensional, and it is difficult to perform a quality inspection because its detection is not easy. In addition, the inspection method has not yet been developed since it is a recently developed product. To solve this problem, we propose a method to calculate the width of pixels for a specific height from the image histogram of a 3D film image and classify it based on a threshold. The proposed algorithm uses the feature that the widths of pixels by height in the image histogram of the good 3D film image are wider than the image histogram of the bad 3D film image. In the experiment, it was confirmed that the position of the height section of the image histogram has the highest classification accuracy. Through comparison tests with conventional algorithms, we showed excellent classification accuracy for 3D film image classification. We verified that it is possible with high accuracy even if the image’s contrast is low and the patterns in the image are not detected.