, ,

Abstract

This paper compares digital image compression in various color spaces using Hybrid Haar wavelet transform. As DCT has high energy compaction property, it is combined with Haar transform generating Haar-DCT hybrid wavelet transform. Different component sizes are used and error at different compression ratios up to 32 is observed. At higher compression ratios 16-16 component pair gives minimum error. This pair is further used to compress images in different color spaces like RGB, KLUV, YUV, YIQ, XYZ and YCbCr. KLUV color space show considerably less Root Mean Square Error (RMSE) and Mean Absolute Error (MAE) than other color spaces. Average Fractional Change in Pixel Value (AFCPV) in KLUV color space is nearly half than other color spaces. RMSE, MAE these measures are not sufficient as it give perceived error to human visual system. Hence other error metric Structural Similarity Index (SSIM) is used to observe performance superiority of various color spaces. SSIM in KLUV and RGB color space is equal having value 0.994 which is closest to one indicating better but similar image quality in both color spaces. This observation cannot be emphasized using traditional error metrics like RMSE, MAE and AFCPV.