Method and system for searching graphic images and videos

The method of clam 1, wherein:

• the step of receiving at least one input specifying an inquiry further comprises:
• defining a text item for a selected image attribute;
• and the step of analyzing the image data further comprises the step of:
• identifying at least one selected image having a text annotation substantially similar to an text item defined for the selected image attribute.

A computer implemented method for determining if an image contains a selected object, comprising the steps of:

• providing a template for the selected object, the template defining geometrical and statistical image characteristics of the selected object, and having a plurality of regions;
• providing for each region of the template a property rule defining a threshold value for the image characteristic in the region of the template;
• applying the template the image to produce a plurality of image regions corresponding to the template regions;
• for each image region, determining a value for the image characteristic;
• scoring the image according to the image characteristic values for all image regions to determine an image score;
• providing a threshold image score representing a minimum score for an image known to contain the object;
• comparing the image score to the threshold;
• and
• designating the image as containing the object if the image score exceeds the threshold.

A system for searching images, comprising:

• a storage device containing a plurality of images, and a plurality of icons representing image attributes;
• a display coupled to the storage device for displaying the images and the icon;
• a low level analyzer coupled to the storage device and retrieving therefrom selected images to produce from each selected image, image data, and storing the image data in the storage device;
• a user interface controller coupled to a display and the storage device, and further coupled to an input device, the user interface controller retrieving the icons from the storage device and displaying the icons on the display, the user interface controller accepting signals from the input device designating selected icons on the display, and storing the image attributes represented by the selected icons in association with each other in the storage device to form a search inquiry;
• a high level analyzer coupled to the storage device and retrieving therefrom the search inquiry and the image data, the high level analyzer comparing each image attribute in the search inquiry with selected image data to determine a relative measure of similarity between the image data and the image attribute, the high level analyzer retrieving a selected number of images substantially similar to the image attributes of the search inquiry.

A method for determining whether an input image of indeterminate scale contains a selected image portion having a fixed scale, the method comprising the steps of:

• a) providing a set of eigen-images E, each eigen-image E.sub.i having an weighting factor w.sub.i, such that the image portion is represented by a combination of all the eigen-images according to their respective weighting factors;
• b) projecting the input image onto each eigen-image to produce a projected weighting factor for each eigen-image;
• c) combining the eigen-images according to their respective projected weighting factors to produce a reconstructed input image;
• d) comparing the reconstructed input image with the image portion to produce a reconstruction error;
• e) resampling the input image according to the reconstruction error;
• f) repeating steps b) through e) until the reconstruction error is minimized;
• and
• g) comparing the minimized reconstruction error with a threshold error associated with an image known to contain the image portion, wherein the input image contains the image portion if the minimized reconstruction error is less than the threshold error.