Rearranged Radon Transform Based Noise Robustness Image Retrieval

Abstract

This study proposed a new image retrieval algorithm in which the existing Radon transform which was used for shape retrieval is reinforced with noise invariance. For this, a Radon transform was performed on an inquiry image which had been preprocessed to extract vector values and then the vector values were arranged depending on size to extract a second feature vector. After clustering and normalizing the levels of vector values based on the second feature vector, the feature vector was created. For a simulation on the noise robustness of the image retrieval system proposed, diverse images were used in this experiment. For performance analysis, the system proposed was compared with the retrieval system using the general Radon transform. As a result, the image retrieval system with noise robustness was between two and three times more robust to geometric transforms such as turning, expansion and scaling-down presented in the retrieval system using the general Radon transform.