Noise Invariant Content Based image retrieval using CDH

This paper presents a noise invariant image analysis using image histograms which can be used for retrieval of similar images. Noise can be defined as degradation in the pixel values. The sources of noise in digital images arise during either image acquisition or transmission. Therefore, robust noise invariant analysis and comparison of images is important for digital image libraries and multimedia databases. The paper uses a form of Histogram known as Color Difference Histogram (CDH) for Content Based Image Retrieval. The main essence of CDH is that it takes into account the perceptually uniform color difference between two points under different background in accordance with the color and the edge orientations. Results show that CDH is noise invariant upto 40% of Gaussian and 50% of Uniform noise. It gives accurate results even if the query image is degraded by additive noise components. KeywordsNoise models; image retrieval; Gaussian noise; Uniform noise; CDH