|M.Sc Student||Emilia Lavi )Burlak(|
|Subject||Correlation vs. De-correlation of Color Components in Image|
|Department||Department of Electrical Engineering||Supervisor||Professor Porat Moshe|
|Full Thesis text|
It is well known that spectral components of natural images, such as RGB, are highly correlated. In this work we have explored the inter-color correlation characteristics of color images, their relation to main statistical properties of the image and their joint behavior in several color planes. Most image compression algorithms tend to decorrelate the color components as part of the compression process. A typical example of the decorrelation approach is the baseline JPEG algorithm, applying YUV color transforms to reduce spectral redundancy and enable low rate chrominance encoding. However, while observing the YUV components as three separate images, considerable resemblance is noted, implying that substantial mutual spectral information has remained and has not been exploited to reduce the bit-rate. To improve the compression, new approaches to color image compression are introduced, taking advantage of the high inter-color correlation to perform efficient encoding of spectral information. The first approach enhances the inter-color correlation of an image prior to compression to provide improved approximation of two of the components using the third one. The second approach employs DCT block transform to allow high quality encoding of spectral information that the human visual system is more sensitive to. Our conclusion is that exploitation of mutual spectral information, as proposed, can improve the coding of chrominance information in color image compression.