LOCAL IMAGE SIMILARITY MEASUREMENT METHOD BASED ON L1 INTERVAL SCALE, WHICH AUTOMATICALLY ESTIMATES A THRESHOLD VALUE IN A RANDOM DETERIORATION LEVEL ABOUT A SIMILARITY STANDARDS OF A RADOM SIZE BASED ON L1 DISTANCES

摘要

PURPOSE: A local image similarity measurement method based on L1 interval scale is provided to be applied to image recovery applications such as a computer vision, machine learning, and super resolution, in-painting, texture synthesis, segment, and object/scene/texture categorization, and other embodies.;CONSTITUTION: An imaging condition is obtained. Based on factors, a proper patch size is determined. From a LUT(Lookup Table), a corresponding threshold value is selected. Threshold value items in the LUT is based on branch elements including smooth conversion between desired similarity ratio, the imaging conditions and various patch size embodiments. A similarity of a local image is measured. A local image similarity standard can be used in applications including restoration, classification, segmentation and detection.;COPYRIGHT KIPO 2011