Industrial Technology Research Institute

Tai-Hui Huang

Wen-Hao Wang

Ming-Yu Shih

Tsung-Chan Li

An automatic image classification method applying an unsupervised adaptation method is provided. A plurality of non-manually-labeled observation data are grouped into a plurality of groups. A respectively hypothesis label is set to each of the groups according to a classifier. It is determined whether each member of the observation data in each of the groups is suitable for adjusting the classifier according to the hypothesis label, and the non-manually-labeled observation data which are determined as being suitable for adjusting the classifier are set as a plurality of adaptation data. The classifier is updated according to the hypothesis label and the adaptation data. The observation data are classified according to the updated classifier.

Unsupervised adaptation method and automatic image classification method applying the same

A moving object detection method and an image processing system thereof are provided. First, a pixel-wise distance of a received image to a reference image is computed to obtain a distance map. A histogram analysis is performed on the distance map to obtain a distance distribution. An entropy value of the distance distribution is computed and a peak distance value which is with a maximum occurrence probability in the distance distribution is searched out. Then, by using a mapping rule, the entropy value and the peak distance value are transformed into a decision threshold value. The decision threshold value is applied in classifying the pixels of the distance map into a group of foreground attributes and a group of background attributes and thereby moving objects in the current image are obtained.

Moving object detection method and image processing system for moving object detection

A foreground image separation method is disclosed to separate dynamic foreground and static background in a sequence of input images which have been processed either with automatic white balance or brightness control by the camera. First, an input image is received from the camera, and then a white balance or brightness compensation is performed to the input image according to a reference image to generate a compensated image with background color and background brightness which are approximately similar to that of the reference image. Finally, a background subtraction algorithm is performed to the compensated image to generate a background separation result. The background subtraction algorithm could be a Gaussian Mixture Model based algorithm. The method could process successive images received from the camera to continuously generate background separation results and update the reference image accordingly, such that video surveillance system could adapt to the change of illumination.

Growing community of inventors

Yunlin County, Taiwan

Tai-Hui Huang