Implementation of Weighted Sample Pixel method for image retrieval application

Color feature based image retrieval methods using color histogram is one of the most frequently used method, mainly because it produces high retrieval results. The weakness of existing histogram methods are their retrieval inaccuracy, caused by similarity in histogram feature even for different images. This research is attempting to develop and implement a slightly different approach by proposing an experimental method which is called Weighted Sample Pixel (WSP). As the name suggested, the method will define the feature of an image based on the weighted pixel which selected as samples taken from the image. This is why the method has no longer need for any histogram assist to define the image unique features. The result of this research is an image retrieval application which successfully implemented the WSP method. The retrieval result is an image taken from the image database which has the exact same feature as the one which served as the base image to be searched. Retrieval average rate is 100% when the method is applied to the application. Sample point quantities and their sizes had a great affect on the retrieval time used.