Median and Laplacian Filters based Feature Analysis for Content Based Image Retrieval Using Color Histogram Refinement Method

Median and Laplacian filters are used to remove noise from images but some mount of information is also lost. Edge extraction and sharpening methods are used to restore the information lost by median and Laplacian filters. Histogram is used to extract features from filtered image but it has problem that images with diverse appearance will have the same histograms because the spatial information in image does not preserve. To preserve spatial information, we quantize histograms into bins. In each bin the statistical features are calculated using the spatial information of regions. For similarity Sum-of-Absolute Differences (SAD) is used to calculate distance between query and database images. Retrieved images are displayed according to the optimized threshold value of the percentage of maximum of distance values. Experiments on the Corel database give results which show that the statistical features of histogram using spatial information are robust in retrieval of images based on Laplacian filter.