Image compression using hybrid wavelet transform with varying proportions of cosine, Haar, Walsh and Kekre transforms with assorted color spaces

Due to rapid growth in technology, billions of images are captured and shared through internet every hour. Lossy image compression techniques minimize the size of the image without degrading the visual quality of the image. Previous work has proved that the hybrid wavelet transforms (HWT) are better than the respective constituent orthogonal transforms in image compression when considered in equal and varying proportions [1, 2, 3]. Here the extended performance comparison of HWT for image compression is presented with assorted color spaces, varying the constituent transforms and the proportions of the constituent transforms to test the effect on quality of image compression. The experimentation is done on set of 15 images by varying the constituent transforms, proportion of constituent transforms, compression ratios (CR) and color spaces. The constituent transforms used to generate HWT are Cosine transform, Kekre transform, Walsh transform and Haar transform. Here five proportions of constituent transforms alias 1:16, 1:4, 1:1, 4:1, 16:1 are considered for generation of HWT. Results prove that 4:1, 1:1, 1:4 proportions of constituent transforms in HWT gives better performance as compared to all other proportions of constituent transforms for different compression ratios. For 95% compression ratio, 4:1 ratio of Cosine- Haar constituent transforms in HWT with LUV color space gives better results. 1:1 ratio of Cosine-Haar constituent transforms give better results for compression ratios between 75% and 90% with RGB color space. For lower compression ratios, 1:4 proportions of Cosine-Haar constituent transforms in HWT gives better performance with RGB color space.