| Summary: | 碩士 === 朝陽科技大學 === 資訊管理系碩士班 === 92 === Due to the rapid evolution of multimedia and Internet, increasingly digital images are being captured and stored. Hence, how to develop a fast and accuracy image retrieval system is very importance.
Color histogram is one of the most often used features in color-based image retrieval. The advantage of the color histogram is simple and easy computation. This feature is also indifferent to image size variant and rotation variant of objects in image. However, it can describe only the global properties of an image, not the local properties, such as it cannot explain the areas of objects and the spatial relationship between objects in an image.
To deal with above problems, this thesis proposes two features, object color-area feature (OCA feature) and object color-spatial feature (OCS feature). The OCA feature combines the colors with areas of the objects in an image; meanwhile the OCS feature integrates the object colors and the spatial relationships between objects. Based on the OCA and OCS features, this thesis provides two image retrieval systems —object color-area based image retrieval system (OCA-Based system) and object color-spatial based image retrieval system (OCS-Based system). Both systems are insusceptible to the image size variant and rotation variant of objects in images. The experimental results also show that these two systems provide good efficiencies in storage space and accurate rate for retrieving the desired database images.
|
|---|
