Quantitative performance measures of color transformations in hybrid vector scalar quantization

In this digital world, increasing demand for images with higher resolution and size has led to a need for more effective methods of compression. This is especially time for color images where there is three times as much data as grayscale images. This thesis discusses the fundamentals of color image processing techniques, with a particular focus on color space transformations and applications, as well as image compression via JPEG, JPEG 2000, and Hybrid Vector Scalar Quantization Compression (HVSQ) methods. The objective is to utilize color spaces to optimize HVSQ compression, thereby enabling it to supersede such compression standards as JPEG and JPEG 2000. So in order to further this goal, comparisons were made between HVSQ and the JPEG standards. Results were compared using the following calculations: Peak signal to Noise Ratio (PSNR), entropy, and Q-Index. Multiple types of images were used in the study, in order to verify the robustness of the color spaces and HVSQ performance. The ten test images that were used can be broken into the following classes of images: computer generated, natural, high frequency, medical, high intensity, and human images. Additionally, multiple bit rates were examined, ranging from 10 to 100, in increments of 10. In general, the reversible YCbCr space was found to have the best performance, allowing HVSQ to outperform JPEG 2000 in most cases.